1. A vision based rapid, non-destructive, and automated method for grading chili quality comprises of an image data collection (1), pre-processing, normalization mechanism, segmentation, feature extraction (2), wireless based sensors systems, machine vision learning, neural network systems, real-time data acquisition from chilli samples (1), measuring the spectrophotometer color values (8), on-site computation, Neural Compute Stick, On-Prem device (5) and an optimized CNN model; said optimized CNN model, hosted in an edge device, grades/classifies chilli samples and said feature extraction involves capturing of details related to the color, texture, and statistical parameters of the chilli samples (4).

2. An automated method for grading chili quality as claimed in claim 1, wherein said Convolutional Neural Network model is trained (3) for chilli grading using pre-processed labelled datasets using precise color, texture, and statistical features.

3. An automated method for grading chili quality as claimed in claim 1, wherein said an image data collection is done by employing wireless sensor-based systems for efficient collection of chilli sample images and real-time data acquisition (2); said pre-processing phase ensures and enhances data quality of the images received as input from image data collection module.

4. An automated method for grading chili quality as claimed in claim 1, wherein said feature extraction involves comprehensive feature extraction methods, encompassing color, texture, and statistical parameters to provide nuanced evaluation of chilli samples.

5. An automated method for grading chili quality as claimed in claim 1, wherein said flexibility in deployment is achieved by hosting the classification model on both cloud platforms as well as On-Prem devices.

6. An automated method for grading chili quality as claimed in claim 5, wherein said the chilli grading/classification model hosted on an On-Prem Device (3) comprising of: a) training of model, b) profiling, tuning, compiling and, c) prototyping.

7. An automated method for grading chili quality as claimed in claim 1, wherein said measuring the spectrophotometer color Values samples (8), comprising factors, L*, representing the lightness on a scale of 0 to 100; a* and b*, representing the chromaticity of the samples; and DE*, representing the total color difference value.

8. An automated method for grading chili quality as claimed in claim 1, wherein said method for comprehensively evaluating chilli samples, comprising: h) applying deep learning methods within a Convolutional Neural Network (CNN) model; i) the optimized CNN model is deployed on a Neural Compute Stick edge device; j) extracting features from the chilli samples, including color, texture, and statistical parameters; k) training the CNN model to recognize and analyze the extracted features; l) treating a comprehensive dataset for nuanced evaluation of the chilli samples based on the analyzed features; m) the user can test the query chilli image sample based on chilli inference model on edge device and output the chilli quality; n) the CNN model’s adaptability allows it to learn and adjust to new chilli varieties, ensuring its widespread effectiveness and versatility.

9. An automated method for grading chili quality as claimed in claim 1, wherein said an edge device for real-time monitoring and analysis, comprising: d) an optimized convolutional neural network (CNN) model for on-site computation; e) a Neural Compute Stick (301) integrated on the chip, configured as a neural network accelerator for managing deep learning applications, and; f) delivering 1 TFLOPS of performance for a small package that fits into hand-held device and delivers high- performance even in severely power-constrained environments.

10. An automated method for grading chili quality as claimed in claim 1, wherein said method involves steps of: k) an input image, involves a module for efficient collection of chilli sample images; l) employing wireless sensor-based systems for real-time data acquisition; m) a pre-processing phase ensuring data quality; n) normalization & segmentation ensures consistent and standardized input, and enhances the delineation of chilli features respectively, contributing to the accuracy of the grading process; o) feature extraction methods, encompassing color, texture, and statistical parameters to provide a nuanced evaluation of chilli samples; p) training of labelled datasets; q) a database of labelled training images to train the model on a Cluster; r) make it available as a trained network for remotely operated devices; s) chilli grading/classification model hosted on laptops, cloud platforms or an On-Prem device (5); t) a three-grade classification system, distinguishing chilli samples into Chilli Grade 1 (Good), Chilli Grade 2 (Medium) and Chilli Grade 3 (Defect) categories based on the features, encompassing color, texture and statistical parameters.

Description:Field of the Invention

The invention relates to the field of agricultural technology and quality assessment. More particularly the invention provides a methodology for Chilli quality identification and grading. Even more particularly the invention provides rapid, non-destructive, and Automated method for Grading Chilli Quality and Colour Count using IOT & edge analytics.

Background of the Invention

In the agricultural landscape, the grading of chilli produce has long been a manual and time-consuming process, fraught with subjectivity and inefficiencies. Traditional methods involve manual inspection or chemical analyses, leading to inconsistencies and delays in decision making for farmers and industries alike. Recognizing these challenges, present invention emerges as a response to the critical need for a technologically advanced solution that streamlines the chilli grading process.

In the past, chilli grading relied on human judgement, resulting in subjective evaluations that varied across individuals. Additionally, chemical method such as High-Performance Liquid Chromatography (HPLC) or Photo spectrometer analysis were employed, necessitating time-consuming processes and sending samples to food labs for testing. This manual and chemical centric approach not only proved inefficient but also posed challenges in maintaining the quality of oleoresin, a vital component in various industries like cosmetics, food and pharmaceuticals. Consequently, there emerged a compelling necessity for a non-destructive, real-time grading system that could accurately grade chilli quality. Many researchers and inventors have worked on it and developed some related works but they are lacking in design and missing some features. Some of such inventions are discussed below.

Reference has been made to CN110479636B titled “Method and device for automatically sorting tobacco leaves based on neural network” by SHENZHEN WEILAN INTELLIGENT TECH CO LTD dated
19/07/2019 which provides a method and a device for automatically sorting tobacco leaves based on a neural network, wherein the method comprises the following steps: collecting natural light pictures of training tobacco leaves, and analyzing the natural light pictures according to preset tobacco leaf characteristic dimension information to obtain natural characteristic information data of the training tobacco leaves; carrying out spectral analysis on the training tobacco leaves to obtain spectral characteristic information data of the training tobacco leaves; corresponding natural characteristic information data and spectral characteristic information data of the trained tobacco leaves, and carrying out neural network training to obtain a correlation model of the natural characteristics and the spectral characteristics of the tobacco leaves; establishing a spectrum tobacco leaf grading strategy of tobacco leaves according to the spectrum characteristic information data, and obtaining a natural grading strategy of the tobacco leaves based on the association model; acquiring a natural image of the tobacco leaves to be sorted, analyzing to obtain natural characteristic information data of the tobacco leaves to be sorted, comparing with a natural grading strategy to obtain a sorting grade of the tobacco leaves to be sorted, and sorting the tobacco leaves to be sorted according to the sorting grade. The invention realizes intelligent, efficient, standardized and low-cost tobacco leaf sorting.

Another reference has been made to JP2006239602 titled “GRADING METHOD FOR FRUIT AND VEGETABLE” by UNIV HIROSAKI dated
04/03/2005 which discloses a grading method for fruits and vegetables for determining their grades by their images obtained by photographing fruits and vegetables has the process of making a two- dimensional color characterization maps peculiar to the kinds of fruit

and vegetables by self-organization by an interconnecting neural network using data converted to the three primary colors of red (R), green (G) and blue (B) obtained from the respective elements of the images photographing fruits and vegetables to be subjected to the grade determination, and an HSV system, and the process of outputting the grades of the desired fruit and vegetables by deciding a plurality of amounts of two-dimensional characterization by samples obtained by sorting the fruits and vegetables to the grades from the color/position information of the object materials which are obtained using the two-dimensional characterization maps and labelling, and by acting an N layer neural network to the amounts of two-dimensional characterization for the grades.

Another reference has been made to CN115953731 titled “Intelligent coal flow monitoring data analysis method for improving CNN algorithm model” by Henan Baojia Information Technology Co ltd dated 14/12/2022 which discloses an intelligent coal flow monitoring data analysis method for improving a CNN algorithm model, which is applied to the field of coal flow monitoring or information processing; the technical problem to be solved is analysis of coal flow monitoring data, and the adopted technical scheme is an intelligent coal flow monitoring data analysis method for improving a CNN algorithm model, which comprises the following steps that (S1) a dome camera in a monitoring area acquires coal flow monitoring data by adopting a machine vision technology; (S2) denoising and eliminating the collected coal flow monitoring image by an improved dark channel prior filtering method; (S3) the coal flow monitoring image subjected to denoising processing is transmitted to a data storage center through a network in an indirect mode, and an improved CNN algorithm model is adopted to carry out data analysis on the coal flow monitoring image; (S4) transmitting the abnormal analysis result to a monitoring alarm module, and displaying the abnormal analysis result through an

OLED display; the invention can reduce the interference of noise to the coal flow monitoring image and improve the accuracy of analyzing the abnormal condition of the coal flow monitoring image.

Another reference has been made to KR102578920 titled “Apparatus for PET sorting based on artificial intelligence” by NHC Co., Ltd dated 16/11/2021 which relates to a PET screening device based on artificial intelligence. The PET sorting device according to the present invention includes a learning unit that trains a pre-built learning model that learns to select the color of plastic waste included in a waste image of plastic waste, and a vision sensor to capture a plurality of plastic wastes input to the conveyor. An image acquisition unit that acquires a waste image, an analysis unit that inputs the acquired waste image into a learned model to extract location information on plastic waste of a preset specific color, and an analysis unit that extracts location information about plastic waste of a preset specific color. A spectrum acquisition unit that sets a light irradiation area using location information, transmits the set light irradiation area to a spectral sensor, and then receives the spectrum of plastic waste located in the light irradiation area from the spectral sensor, the acquired spectrum a determination unit that determines whether the plastic waste located in the light irradiation area corresponds to PET, and a control unit that sets the location information on the plastic waste determined to be PET as the final sorting coordinates and transmits it to the pickup robot. Includes. According to the present invention, by providing a total platform that can automatically sort and separate recycled PET, the sorting process in the recycling field can be automated and the sorting accuracy and sorting speed can be improved.

Another reference has been made to IN202341086759 titled “An automatic processing of medical images using segmented K-means clustering algorithm (SKCA) based” by M. JAWAHAR dated 19/12/2023 which discloses an approach to achieve accurate results in medical

Image segmentation, two methods are combined this study The three clustering strategies include over-segmentation, post-segmentation merging on the original partitions to Minimize the number of erroneous edges, and unsupervised learning algorithms. We demonstrated that the segmentation maps generated by out suggested approach had less partitions than the segmentation maps generated by the current methodology. Keywords: SKCA, segmentation and unsupervised learning.

Another reference has been made to US10918016B2 titled “System and method for grading agricultural commodity” by INTELLO LABS PRIVATE LTD [IN] dated 06/04/2018 which discloses a system and method for grading agricultural commodity comprising a moisture sensor, a grain apparatus, a conveyor belt driven by motor, a camera to take plurality of images of the agricultural commodity uniformly distributed on the conveyor belt, a communication device for transmitting information to a processor, which processes the image to ascertain the quality parameters of the agricultural commodity and compares with the standards used for grading, a user interface for displaying the information of grading and a printer to print the information of grading. Further, the moisture sensor provides the moisture reading of a sample of agricultural commodity put through the grain apparatus. Furthermore, the parameters are analysed using image processing technique and compared with the Standard used for grading of the agricultural commodity to ascertain the grading. Additionally, the system can be integrated or several components at different places connected through network.

Another reference has been made to US2020273172A1 titled “Crop grading via deep learning” by IBM [US] dated 27/02/2019 which discloses methods and systems for crop grading and crop management. One or more images of crops are obtained and one or more crop related features are at least one of identified or extracted

from the one or more images. A crop health status is determined based on the one or more crop related features, an environmental context, a growth stage of the crop, and a farm cohort by using a computerized deep learning system to perform an automated growth stage analysis. One or more actions are at least one of recommended, triggered, and performed.

Another reference has been made to CN115690582A titled “Winter wheat identification method based on hyperspectral data” by UNIV CHUZHOU dated 31/10/2022 which provides a winter wheat identification method based on hyperspectral data, and belongs to the technical field of crop identification, and the method comprises the steps: constructing a decision tree classification model based on knowledge rules, which comprises the steps: collecting the hyperspectral data of winter wheat in a growth period, and taking the hyperspectral data as original hyperspectral data; randomly selecting a part of samples from the original hyperspectral data as training samples; performing first-order differential transformation on the training sample to obtain a first-order differential spectrum; according to the original hyperspectral data and the first-order differential spectrum, extracting spectral characteristics of the winter wheat in different growth periods and different growth Vigors; setting a decision tree classification model to identify selection conditions of the winter wheat according to spectral characteristics of different growth periods and different growth Vigors of the winter wheat, and completing construction of the decision tree classification model; and inputting the hyperspectral data of a certain crop growth period into the decision tree classification model to obtain an identification result of whether the crop is winter wheat or not. The method can be used for identifying the winter wheat

Another reference has been made to CN106706546A titled “Analysis method for artificial intelligence learning materials on the basis of

infrared and Raman spectrum data” by ZHONGSHAN CITY TENGCHUANG TRADE CO LTD dated 28/12/2016 which provides a rapid analysis and processing method for infrared and Raman spectrum data on the basis of artificial intelligence. The rapid analysis and processing method comprises the following steps of: acquiring an infrared and Raman spectrum library of a single-sample learning group; carrying out statistical filtering on data of the spectrum library to remove suspect and invalid data; carrying out unsupervised learning classification on the spectrum library of the learning group by artificial intelligence; carrying out chemical analysis, verification and classification on every type of samples distinguished by classification results; if a sampling result cannot meet the application needs, returning an artificial intelligence program and carrying out supervised learning classification by using chemical analysis results; by multiple repeated classification and iteration, using the artificial intelligence program to generate an analysis prediction model capable of meeting the application needs automatically; and using the model for rapid spectrum analysis and detection of unknown samples in the type of samples. The rapid analysis and processing method provided by the invention is applicable to rapid spectrum analysis and detection of a large number of samples, can realize accuracy adjustment according to the needs, and can be rapidly applied in different sample testing scenes, so that the application prospect and potential are great.

Another reference has been made to IN202321019121 titled “DEVELOPMENT OF NON-INVASIVE VISIBLE LIGHT MULTISPECTRAL IMAGING TECHNIQUE FOR GRADING AND QUALITY EVALUATION OF
APPLE FRUIT” by Symbiosis International (Deemed University) dated 21/03/2023 which relates to a system for determining the type and class of apple fruit. The system comprises a combination of existing components, including an RGB LED ring (7), a camera, and a wooden box (5) capable of accommodating different size fruits. The system

also 5 incorporates a deep learning algorithm, named "Apple Net", which is trained to classify apples into different categories based on their characteristics. The device includes a USB camera and an active illumination source capable of emitting visible spectral wavelengths. The camera captures images of the apples, which are then processed while enhancing features of interest. The system also includes a unique code that stitches the captured image with a resolution of 10 17280x1080x3 and then rescales it to one-tenth of its original image size. This image is then fed to the Apple Net algorithm, which classifies the apples into three different categories, including Red Delicious, Alpita, and Royal Gala.

Another reference has been made to US11195015B2 titled “IOT BASED FARMING AND PLANT GROWTH ECOSYSTEM” by Bao Tran & Ha
Tran dated 12/07/2021 which discloses an agricultural method that includes providing a positive air pressure chamber to prevent outside contaminants from entering the chamber; growing crops in a plurality of cells in the chamber, each cell having multi-grow benches or levels, each cell further having connectors to vertical hoists for vertical movements in the chamber; maintaining pre-set temperature, humidity, carbon dioxide, watering and lighting levels to achieve predetermined plant growth; using motorized transport rails to deliver benches for operations including seeding, harvesting, grow media recovery, and bench wash; dispensing seeds in the cell with a mechanical seeder coupled to the transport rails; growing the crops with computer controlled nutrients, light and air level; and harvesting the crops and delivering the harvested crop at a selected outlet of the chamber.

Another reference has been made to WO/2020/227429 titled “Platform for facilitating development of intelligence in an industrial internet of things system” by STRONG FORCE IOT PORTFOLIO 2016, LLC [US]/[US] dated 02/12/2019 which provides a platform for

facilitating development of intelligence in an Industrial Internet of Things (IIoT) system can comprise a plurality of distinct data-handling layers. The plurality of distinct data-handling layers can comprise an industrial monitoring systems layer that collects data from or about a plurality of industrial entities in the IIoT system; an industrial entity- oriented data storage systems layer that stores the data collected by the industrial monitoring systems layer; an adaptive intelligent systems layer that facilitates the coordinated development and deployment of intelligent systems in the IIoT system; and an industrial management application platform layer that includes a plurality of applications and that manages the platform in a common application environment. The adaptive intelligent systems layer can include a robotic process automation system that develops and deploys automation capabilities for one or more of the pluralities of industrial entities in the IIoT system.

Another reference has been made to US2022384043A1 titled “System and methods for enhanced photodetection spectroscopy using data fusion and machine learning” by LIGHTSENSE TECH INC [US] dated 28/05/2021 which relates to a method for detection of pathogens, biomarkers, or any compound using data fusion and machine learning. The method includes generating, with a first miniature UV absorption spectrometer of a multi-spectral optical device, a first absorption spectral output based on receiving an absorbance light channel from a sample, generating, with a second miniature UV fluorescence spectrometer of the multi-spectral optical device, a second emission spectral output based on receiving an emission light channel from the sample and performing, with the multi- spectral optical device, data fusion between the first absorption spectral output and the second emission spectral output to generate fused data.

Another reference has been made to US2018084741A1 titled “Method and system for industrial internet of things data collection for a chemical production process” by STRONGFORCE IOT PORTFOLIO 2016 LLC [US] dated 09/05/2016 which relates to a system, method and apparatus for data collection in an industrial production environment are described. The system may include, a data acquisition circuit structured to interpret a plurality of detection values, each of the plurality of detection values corresponding to input received from at least one of a plurality of input sensors which includes a detection package, each of the plurality of input sensors operatively coupled to at least one of a plurality of components of an industrial production process, a data analysis circuit structured to analyse a subset of the plurality of detection values to determine a sensor performance value of at least one of the plurality of input sensors, and an analysis response circuit structured to adjust at least one of a sensor scaling value or a sensor sampling frequency value, in response to the sensor performance value.

Another reference has been made to WO2014078862A1 titled “Blending of agricultural products via hyperspectral imaging and analysis” by ALTRIA CLIENT SERVICES INC [US] dated 19/11/2012 which provides a method for blending of agricultural product utilizing hyperspectral imaging. At least one region along a sample of agricultural product is scanned using at least one light source of different wavelengths. Hyperspectral images are generated from the at least one region. A spectral fingerprint for the sample of agricultural product is formed from the hyperspectral images. A plurality of samples of agricultural product is blended based on the spectral fingerprints of the samples according to parameters determined by executing a blending algorithm.

Another reference has been made to WO2022175309A1 titled “Methods for analysing plant material, for determining plant material

components and for detecting plant disease in plant material” by KWS SAAT SE & CO KGAA [DE] dated 17/02/2021 which relates to a method for analysing a crop sample comprising a target plant material with soil tare adhered thereto, in particular soiled plant material. Further, the invention relates to a method for generating first calibration data for analysing a crop sample comprising a target plant material with soil tare adhered thereto, and an analysis assembly for analysing a crop sample comprising a target plant material with soil tare adhered thereto. In addition, the invention relates to an arrangement for analysing a crop sample comprising a target plant material with soil tare adhered thereto, a sugar production facility, and the use of an analysis assembly in a sugar production facility. Further, the invention relates to a method for determining components in sugar beets for sugar production. Further, the invention relates to an arrangement for determining components in sugar beets, and to a sugar production facility. Further, the invention relates to a method for generating calibration data for the determination of components in sugar beets, and to a use of an analysis assembly and/or an arrangement and/or a method. Further, the invention relates to a method for detecting plant diseases in plant material and/or physiological properties influenceable by environmental stress in plant material, an analysis assembly for detecting plant diseases in plant material, an arrangement for detecting plant diseases in plant material, and a control unit for controlling an analysis assembly and/or for receiving data from an analysis assembly.

Another reference has been made to WO2011027315A1 titled “Grading of agricultural products via hyper spectral imaging and analysis” by MOSHE DANNY S [IL]; DANTE HENRY M [US]; DEEVI SEETHARAMA C [US]; HINTON CURTIS M [US] dated 04/09/2009
which relates to a system (10) for grading an agricultural product (P) employing hyper-spectral imaging and analysis. The system includes at

least one light source (12) for providing a beam of light, an interferometer or a prism array for dispersing electromagnetic radiation emitted from said agricultural product (P) into a corresponding spectral image, a light measuring device for detecting component wavelengths within the corresponding spectral image and a processor (34) operable to compare the detected component wavelengths to a database of previously graded agricultural products to identify and select a grade for the agricultural product. A method for grading an agricultural product via hyper-spectral imaging and analysis is also provided.

Another reference has been made to CN110681610A titled “Material Visual Recognizing and Sorting System Based on Convolution Neural Network and Depth Learning” by TANG XIAOQING dated 30/10/2019 which discloses a material visual recognizing and sorting system based on convolution neural network and depth learning. The system comprises a material grading sieve, a distributing device, at first conveying belt, a visual recognizing device, a GPU server, a control device and an action executing device. The material grading sieve is located at the head end of the first conveying belt. The distributing device and the visual recognizing device are sequentially arranged on the first conveying belt in the material conveying direction. The action executing device is arranged at the tail end of the first conveying belt. The vision recognizing device is connected with the control device through the GPU. The visual recognizing device rapidly collects feature information and space position information of materials on the first conveying belt, and transmits the information to the GPU server. The control device is connected with the action executing device and controls the action executing device to act. According to the system, by means of convolution neural network and depth learning, the materials are visually recognized and sorted, the overall efficiency is high, and the intelligent degree is high.

Another reference has been made to WO2023084543A1 titled “System and method for leveraging neural network-based hybrid feature extraction model for grain quality analysis” by WAYCOOL FOODS AND PRODUCTS PRIVATE LTD [IN] dated 12/11/2021 which
relates to a system and method for analysing quality measures related to grains using image classification convolutional neural networks- based prediction model to provide real time price prediction based on defects related to grain size, count and other quality parameters. The system comprises: a plurality of databases, a package holding module, a sensor unit, an image identification unit interconnected to the package holding module and a deep learning-based prediction model with a plurality of database systems via a cloud-based network. The method comprises obtaining images, processing, and storing the images in the database system, receiving and analysing an input image by a deep learning model, selecting appropriate model based on the processed image, generating prediction output based on the selected model in a form of a grain type and combining the sensor input and the prediction output generated by the sensor unit and prediction model.

Another reference has been made to CN113744075A titled “Agricultural Product Nutritional Quality Grading System Based on Artificial Intelligence” by INSTITUTE OF FOOD AND NUTRITION DEVELOPMENT MINI OF AGRICULTURE AND RURAL AFFAIRS;
CHEMMIND TECH CO LTD dated 08/09/2021 which provides an agricultural product nutritional quality grading system based on artificial intelligence, and relates to the technical field of artificial intelligence. The system comprises the following steps: S1, detecting agricultural products, and inputting agricultural product detection data; S2, processing the detected data of the agricultural products; wherein the data processing comprises a multispectral instrument data processing system and an omics data analysis system; S3, carrying

out data modelling on the agricultural product nutritional quality basic database; S4, establishing a chemometrics and machine learning algorithm library for grading the agricultural products; and S5, constructing an agricultural product grading system. By the combination level of agricultural product detection and artificial intelligence, deep understanding of the material basis and the action rule of the nutritional quality of the agricultural products is promoted, grading of the agricultural products is promoted, and transformation development of agricultural production in China from survival type food supply to healthy type nutritional quality improvement is promoted.

Another reference has been made to CN112418130A titled “BP neural network-based banana maturity detection method and device” by UNIV SOUTH CHINA AGRICULT dated 30/11/2020 which discloses a BP neural network-based banana maturity detection method and device. The method comprises the steps of obtaining a banana image of to-be- detected banana maturity; detecting the banana image through a BP neural network banana maturity detection model obtained by pre- training, wherein the detection model is obtained by training a BP neural network model through multiple groups of training data, wherein each group of data of the multiple groups of training data comprises banana images of multiple maturity types and maturity grade information corresponding to the banana images; and determining the maturity of bananas in the banana image according to the detection result of the BP neural network banana maturity detection model. Compared with the traditional detection method, the BP neural network-based banana maturity detection model established by the invention can realize non-destructive detection of banana samples, is high in accuracy and short in time consumption, and improves the economic benefits of bananas.

However, none of the invention discussed above provide a vision based rapid, non-destructive, and Automated method for

Grading Chili Quality and Color count which involves image data collection, wireless based sensor systems, chilli database preparation and integrates artificial intelligence, machine vision learning and IoT to create an optimized Convolutional Neural Network (CNN) model. The said method provides the Grading Chili Quality, using vision features like color, texture, shape, and leveraging edge computing to bring computational power to the field, eliminating the need for bulky hardware.

Objective of the Invention

The main objective of the present invention is to provide a vision based rapid, non-destructive, and automated method for grading chilli quality and chilli color count to access dry chilli fruits & oleoresin.

Another objective of the present invention is to automate manual quality grading with scientific color grading on per with ASTA Colour units.

Another objective of the present invention is to develop a standard method as par with international standards and practices to help the sellers (producers, famers & institutes) then buyers (food/pharma/cosmetics/fabric-textile, meat industries, APMC’s- mandi’s, Spices-Boards & Farmer produce organization (FPO’s), Universities/Institutes) in deciding on procurement and inventory processes in national and international markets.

Another objective of the present invention is to deliver high- performance even in severely power-constrained environments like storage & godown facilities.

Another objective of the present invention is to ensure portability and eliminate reliance on bulky hardware.

Another objective of the present invention is to implement a vision- based AI and machine learning mechanism that provides a detailed analysis of chilli samples for the grading system.

Another objective of the present invention is to ensure real-time connectivity and monitoring capabilities for efficient data analysis.

Another objective of the present invention is to ensure model’s adaptability to deploy on both cloud services and On-Prem devices.

Summary of the invention

The present invention provides a vision based rapid, non-destructive, and automated method for grading chilli quality that focuses on automating the chilli grading process, providing accurate and efficient chilli quality assessment. The said method is anchored with advanced technologies like wireless based sensor systems, Artificial Intelligence, Machine vision Learning, neural network systems and deep learning systems optimizing the post-harvest chilli handling process. The key components include an optimized Convolutional Neural Network (CNN) model deployed on an edge device, enabling real-time and non- destructive grading. The method encompasses image data collection, pre-processing, normalization, segmentation. Further, the labelled chilli dataset is trained, followed by the classification of chilli samples into three grades: Grade-1 (Good), Grade-2 (Medium), and Grade-3 (Defect), which further provides nuanced evaluations based on color, color units, texture, and statistical parameters. The method’s advantages extend to its portability, facilitated by edge computing, eliminating the need for bulky hardware. This innovation empowers farmers, traders, and industries with efficient, accurate, and portable chilli grading system, reshaping decision-making and resource management.

Statement of the invention

The present invention relates to a vision based rapid, non-destructive, and automated method for grading chili quality and base for scientific pricing of chilli as commodity for high value addition involving integration of Artificial Intelligence, Machine vision learning, and IoT technologies which facilitates real-time, non-destructive chilli quality grading; the present invention incorporates the deployment of CNN model on an edge device, streamlining the post-harvest process involving image data collection, pre-processing, normalization segmentation and feature extraction, wherein, the CNN model classifies chilli samples into three grades based on color, texture, and statistical features which eradicates the manual grading drawbacks and helps in pricing of raw materials, chilli produce, value-added products & chilli oleoresin trade for national & global trade for revenue estimation, procurement decisions, and efficient inventory management.

Brief description of the figures

Figure 1 shows the chilli grading/classification model hosted on cloud platform.
Figure 2 shows the Chilli Grading/Classification Model comprising of 4 steps elucidating the grading/classification method under which Step
1 includes Input Image, Pre-processing, Normalization & Segmentation, Step 2 includes training of labelled chilli dataset and processing it using grading/classification model hosted on laptop/ server/ On Prem device, Step 3 includes the three grade classification, distinguishing chilli samples into Chilli Grade 1 (Good), Chilli Grade 2 (Medium) and Chilli Grade 3 (Defect) categories.
Figure 3 shows the chilli grading/classification model hosted on an On-Prem Device comprising of a Neural Compute Stick (NCS) edge device 301, Development Host 302 and a Prototyping Host 303.

Figure 4 shows the different varieties of chilli used for the experiment/trials.
Figure 5 show the On-Prem Model-Prototyping Host comprising of Raspberry Pi 4 501, fitted in a case 502, and a 15W USB-C Power supply cord 503.
Figure 6 shows the test results on Prototype Model based on the classification performance of chilli grading.
Figure 7 shows the hue color model features for byadgi chilli varieties with stalk and destalked including semi dry and dry chilli samples.
Figure 8 shows the analysis report of dry red chilli samples.

Detailed description of the Invention

It should be noted that the particular description with features, designs, components, construction, working and embodiments set forth in the specification below are merely exemplary of the wide variety and arrangement of instructions which can be employed with the present invention. The present invention maybe embodied in other specific forms without departing from the spirit or essential characteristics thereof. All the features disclosed in this specification may be replaced by similar other or alternative features performing similar or same or equivalent purposes. Thus, unless expressly stated otherwise, they all are within the scope of present invention. Various modifications or substitutions are also possible without departing from the scope or spirit of the present invention. Therefore, it is to be understood that this specification has been described by way of the most preferred embodiments and for the purposes of illustration and not limitation.

The present invention provides a vision based rapid, non-destructive, and automated method for grading chili quality with integration of Artificial Intelligence, Machine vision learning, and IoT technologies

which facilitates real-time, non-destructive chilli quality and its oleoresin content grading. The present invention incorporates the deployment of an optimized Convolutional Neural Network model (1) on an edge device, that facilitates real-time, non-destructive grading in the field. Moreover, this innovation aims to address the limitations of manual grading and chemical methods by leveraging advanced technologies such as Artificial Intelligence, Machine vision learning, and IoT.
The food industry and value addition industries purchase chilli produce with high-quality colour and chilli oil (Oleoresin). The present invention provides a method that grades quickly and non-destructively the best quality chilli produce with high oleoresin color. The said method for chilli quality grading considers the significant applications of oleoresin, also known as chilli oil because Oleoresin is a natural colorant used in coloring tablets, cosmetics, food, and fabrics in pharmaceutical applications due to restrictions on the use of synthetic chemicals in products intended for human consumption.
The present invention provides a method for image data collection, pre-processing, normalization, segmentation and feature extraction. The optimized CNN model, hosted on an edge device, grades/classifies chilli samples into three grades i.e., Chilli Grade-1 (Good), Chilli Grade-
2 (Medium), and Chilli Grade-3 (Defect). Furthermore, this invention leverages technologies like wireless based sensors systems, machine vision learning, neural network systems and deep learning, enhancing the post-harvest handling of the chilli produce and extending the marketable life of food produce (1).
The present invention provides a method that comprises of a wireless sensor-based image data collection module (2), a normalization mechanism, and an optimized CNN model for classification. The wireless-sensor based module ensures efficient and real-time data acquisition from chilli samples (1).

In the present invention, the technicalities incorporated the application of deep learning methods within the CNN model and it also includes feature extraction, which involves the capturing of details related to the color, texture, and statistical parameters of the chilli samples. The said CNN model is trained to recognize and analyze these features and create a comprehensive dataset for a nuanced evaluation of chilli samples (1 & 4). The use of deep learning methods ensures that the model can adapt and learn from diverse chilli varieties and environmental conditions, enhancing its robustness and reliability.
The present invention involves the deployment of the CNN model on an edge device employing the utilization of edge computing ensuring that computational power is readily available in the field, eliminating the need for the carrying bulky hardware to the fields across geographical locations.
In the present invention IoT technologies have been seamlessly integrated, enabling connectivity for real-time monitoring and analysis. The edge device, equipped with the optimized CNN model, facilitates on-site computation, eliminating the need for carrying bulky hardware. It also includes a Neural Compute Stick (301) which is a neural network accelerator on the chip, that allows managing deep learning applications based on neural networks. It is capable of delivering 1 TFLOPS of performance for a small package that fits into hand-held device and delivers high-performance even in severely power- constrained environments.
In the present invention, the optimized CNN model is deployed on a Neural Compute Stick edge device, where the user can test the query chilli image sample based on chilli inference model on edge device and output the chilli quality which elucidates the user-friendliness and portability of the method. Moreover, the CNN model’s adaptability allows it to learn and adjust to new chilli varieties, ensuring its widespread effectiveness and versatility.

The present invention provides a method that is non-destructive making it advantageous over the traditional manual grading methods as the present invention’s method involves an automated system for conducting evaluations without compromising the integrity of the chilli samples which not only preserves the quality of the produce but also minimized waste, aligning with sustainable agricultural practices.
The present invention provides a rapid, Non-destructive, and Automated system for Grading Chili Quality and color units that comprising the following steps (2):
• Collection of chilli sample images using the wireless sensor- based module;
• Pre-processing techniques are applied to ensure data quality;
• Normalization and segmentation for consistent and standardized inputs;
• Training of labelled chilli dataset and processing it through grading/classification model;
• Classification model categorizes chilli samples into three grades based on extracted features:
o Chilli Grade 1 (Good)
o Chilli Grade 2 (Medium)
o Chilli Grade 3 (Defect)
The Figure 1 discloses the chilli grading/classification model hosted on cloud platform, that involves a database of labelled training images to train the model on a Cluster and make it available as a trained network for remotely operated devices. The figure also shows the results for the input chili samples, which include Guntur, grade 1, and semi-dry varieties.
In the present invention, Figure 2 discloses the Chilli Grading/Classification Model comprising of a 4-step procedure that elucidates the grading/classification method under which Step 1 includes Input Image, which involves a module for efficient collection

of chilli sample images, employing wireless sensor-based systems for real-time data acquisition, Pre-processing, which involves a pre- processing phase ensuring data quality, Normalization & Segmentation which ensures consistent and standardized input, and enhances the delineation of chilli features respectively, contributing to the accuracy of the grading process; Step 2 includes feature extraction methods, encompassing color, texture, and statistical parameters to provide a nuanced evaluation of chilli samples; Step 3: includes a training of labelled datasets and process it further to the grading/classification model hosted on laptops, cloud platforms or an On-Prem device like Raspberry Pi 4 501, offering flexibility in deployment various operational requirements; Step 4 includes a three grade classification system, distinguishing chilli samples into Chilli Grade 1 (Good), Chilli Grade 2 (Medium) and Chilli Grade 3 (Defect) categories based on the features, encompassing color, texture and statistical parameters.
The Figure 3 discloses the chilli grading/classification model hosted on an On-Prem Device comprising of: a) training of model, b) profiling, tuning, compiling c) prototyping. The said chilli grading/classification model hosted on an On-Prem Device 501, a Neural Compute Stick edge device 301 which is a neural network accelerator on the chip and development host 302 are compiled with Neural Compute Stick 301 and Prototyping host 303, which allows it to tackle deep learning applications based on neural networks, capable of delivering 1 TFLOPS of performance for a small package that fits into hand-held device and delivers high-performance even in severely power-constrained environments.
The Figure 4 discloses the different varieties of chilli used in creating the dataset for the classification model. The input image is pre- processed, to ensure the data quality. Normalization and segmentation are incorporated to ensure consistent and standardized input and

enhance the delineation of chilli features respectively, which further proceeds to the training of labelled datasets and incorporation of grading/classification model hosted on laptops, servers or an On-prem device like Raspberry Pi 4 501. The Chilli dataset used in this method includes BN-Bhug Nag, BD-Byadagi-Dhabi, BMD-Byadgi-Med Dhabi, BSP-Byadgi single patta, GN-Guntur, JP-Jalpaneo, JW-Jwala, KS-Kesar, MD-Mandu, RP-Reshma Patta and SM-Surya Mukhi.
The Figure 5 discloses the On Prem Model–Prototyping Host comprising of a Raspberry Pi 4 Module 501, used for hosting the CNN classification model employing the utilization of edge computing ensuring that computational power is readily available in the field, eliminating the need for carrying bulky hardware to the fields; The said On Prem device is enclosed within a touchscreen case 502 to provide a casing to the module, and a 15W USB-C Power Supply cord 503 is available to supply power to the On Prem device.
Figure 6 discloses the test results of different chilli grades (chilli grade 1, 2 & 3) for standard deviation and average classification based on the classification performance of chilli grading.
Figure 7(a) and 7(b) shows the hue color model features for Byadgi chilly varieties with stalk and destalked based on the dryness and moisture content. Figure 7 (a) shows the hue features for Semi-Dry Chilli Samples with 20% moisture, and Figure 7 (b) shows the hue features for Dry Chilli Sampled with 10% moisture. The said chilli samples are Bhug-Nag, Byadagi-Dhabi, Byadgi-Med-Dhabi, Byadgi- single-patta, Guntur, Jalpaneo, Jwala, Kesar, Mandu, Pimento, Reshma- Patta, Sannam and Surya-Mukhi. The factors incorporated in this classification include Hue-Avg, Hue-StdDev and H-Var.
Referring to Figure 8, discloses the analysis report that includes sample name ‘Dry Red Chilli Wise Stalk Samples’ having A.R. No. QC- NS-611 to QC-NS-618, quantity 100 gms and dry red chilli wise stalk samples sealed in polyethene zip lock covers.

The table given below shows the Spectrophotometer Color Values from Evaluation Lab test report comprising of the factors, L*a*b* color units is disclosed, wherein, L* signifies the lightness from black to white on a scale of 0 to 100; a* and b* depicts the chromaticity of the sample. The table includes a DE* factor which represents the total colour difference value.

Table 1: Spectrophotometer Color Values

S.
No. Sample -
ID
QC-NS-ID Name of Test
L*
a*
b*
DE*
1. CWS-92 QC-NS-611 Color 23.11 12.82 11.17 47.47
2. CWS-102 QC-NS-612 Color 20.13 15.88 11.46 51.54
3. CWS-Local QC-NS-614 Color 19.78 12.19 6.75 49.42
4. CDS-92 QC-NS-615 Color 19.85 16.75 7.80 52.76
5. CDS-102 QC-NS-616 Color 16.72 15.31 7.46 53.65
6. CDS-487 QC-NS-617 Color 16.69 15.55 7.51 53.92
7. CDS-Local QC-NS-618 Color 19.10 16.33 8.84 52.78

The above analysis report is merely for illustration purpose and shall not be construed to limit the scope of the invention. The aforementioned parameters will be tested in accordance with the established requirements and specifications.
The present invention provides a vision based rapid, non-destructive, and automated process for grading chili quality with integration of Artificial Intelligence, Machine vision learning, and IoT technologies which facilitates real-time, non-destructive chilli quality grading and an optimized Convolutional Neural Network model is deployed on an edge device; the process encompasses efficient image data collection, pre- processing, normalization and segmentation, followed by a classification based on color, texture, and statistical parameters; the grading process is streamlined through edge computing elevating the

portability by reducing the need for bulky hardware and with the help of a three-grade classification system, namely, Chilli Grade 1 (Good), Chilli Grade 2 (Medium), Chilli Grade 3 (Defect), the model provides a nuanced evaluation of the quality of chilli produce; Further, hosting on both cloud platform and On-Prem devices helps to achieve flexibility in deployment and the integration of IoT and deep learning systems ensures real-time connectivity and enhances adaptability to diverse chilli varieties and environmental conditions.
Further, the present invention provides a method that emphasizes on identifying quickly and non-destructively the best quality chilli produce during post harvesting and revenue approximation. The proposed method incorporates the integration of technologies like wireless sensor-based systems, artificial intelligence, machine vision learning, neural network systems, databases and deep learning system for post- harvest handling of the chilli produce, ensuring the extension of the marketable life of the food produce.
So accordingly, the present invention provides a vision based rapid, non-destructive, and automated method for grading chili quality comprises of an image data collection (1), pre-processing, normalization mechanism, segmentation, feature extraction (2), wireless based sensors systems, machine vision learning, neural network systems, real-time data acquisition from chilli samples (1), measuring the spectrophotometer color values (8), on-site computation, Neural Compute Stick, On-Prem device (5) and an optimized CNN model; said optimized CNN model, hosted in an edge device, grades/classifies chilli samples and said feature extraction involves capturing of details related to the color, texture, and statistical parameters of the chilli samples (4).
In an embodiment, said Convolutional Neural Network model is trained
(3) for chilli grading using pre-processed labelled datasets using precise color, texture, and statistical features.

In another embodiment, said an image data collection is done by employing wireless sensor-based systems for efficient collection of chilli sample images and real-time data acquisition (2); said pre- processing phase ensures and enhances data quality of the images received as input from image data collection module.
In another embodiment, said feature extraction involves comprehensive feature extraction methods, encompassing color, texture, and statistical parameters to provide nuanced evaluation of chilli samples.
In another embodiment, said flexibility in deployment is achieved by hosting the classification model on both cloud platforms as well as On- Prem devices.
In another embodiment, said the chilli grading/classification model hosted on an On-Prem Device (3) comprising of: a) training of model,
b) profiling, tuning, compiling and, c) prototyping.

In another embodiment, said measuring the spectrophotometer color Values samples (8), comprising factors, L*, representing the lightness on a scale of 0 to 100; a* and b*, representing the chromaticity of the samples; and DE*, representing the total color difference value.
In another embodiment, said method for comprehensively evaluating chilli samples, comprising:
a) applying deep learning methods within a Convolutional Neural Network (CNN) model;
b) the optimized CNN model is deployed on a Neural Compute Stick edge device;
c) extracting features from the chilli samples, including color, texture, and statistical parameters;
d) training the CNN model to recognize and analyze the extracted features;

e) treating a comprehensive dataset for nuanced evaluation of the chilli samples based on the analyzed features;
f) the user can test the query chilli image sample based on chilli inference model on edge device and output the chilli quality;
g) the CNN model’s adaptability allows it to learn and adjust to new chilli varieties, ensuring its widespread effectiveness and versatility.
In another embodiment, said an edge device for real-time monitoring and analysis, comprising:
a) an optimized convolutional neural network (CNN) model for on-site computation;
b) a Neural Compute Stick (301) integrated on the chip, configured as a neural network accelerator for managing deep learning applications, and;
c) delivering 1 TFLOPS of performance for a small package that fits into hand-held device and delivers high- performance even in severely power-constrained environments.
In another embodiment, said method involves steps of:

a) an input image, involves a module for efficient collection of chilli sample images;
b) employing wireless sensor-based systems for real-time data acquisition;
c) a pre-processing phase ensuring data quality;

d) normalization & segmentation ensures consistent and standardized input, and enhances the delineation of chilli features respectively, contributing to the accuracy of the grading process;

e) feature extraction methods, encompassing color, texture, and statistical parameters to provide a nuanced evaluation of chilli samples;
f) training of labelled datasets;

g) a database of labelled training images to train the model on a Cluster;
h) make it available as a trained network for remotely operated devices;
i) chilli grading/classification model hosted on laptops, cloud platforms or an On-Prem device (5);
j) a three-grade classification system, distinguishing chilli samples into Chilli Grade 1 (Good), Chilli Grade 2 (Medium) and Chilli Grade 3 (Defect) categories based on the features, encompassing color, texture and statistical parameters.

While particular embodiments of the present invention have been shown and described, it will be obvious to those skilled in the art that changes and modifications may be made without departing from this invention in its broader aspects and, therefore, the aim in the present invention is to cover all such changes and modifications as fall within the true spirit and scope of this invention.

Advantages of the invention:

1. The present invention provides a quick instant, non-destructive and cost-effective technique.
2. The present invention provides an automated method with high accuracy and time-efficiency.

3. The present invention assists industries such as cosmetics, food, and pharmaceuticals in making procurement and inventory decisions.
4. Delivers high-performance even in severely power-constrained environments.
5. Portable and easily accessible.
, Claims:We Claim:

1. A vision based rapid, non-destructive, and automated method for grading chili quality comprises of an image data collection (1), pre-processing, normalization mechanism, segmentation, feature extraction (2), wireless based sensors systems, machine vision learning, neural network systems, real-time data acquisition from chilli samples (1), measuring the spectrophotometer color values (8), on-site computation, Neural Compute Stick, On-Prem device
(5) and an optimized CNN model; said optimized CNN model, hosted in an edge device, grades/classifies chilli samples and said feature extraction involves capturing of details related to the color, texture, and statistical parameters of the chilli samples (4).
2. An automated method for grading chili quality as claimed in claim 1, wherein said Convolutional Neural Network model is trained (3) for chilli grading using pre-processed labelled datasets using precise color, texture, and statistical features.
3. An automated method for grading chili quality as claimed in claim 1, wherein said an image data collection is done by employing wireless sensor-based systems for efficient collection of chilli sample images and real-time data acquisition (2); said pre-processing phase ensures and enhances data quality of the images received as input from image data collection module.
4. An automated method for grading chili quality as claimed in claim 1, wherein said feature extraction involves comprehensive feature extraction methods, encompassing color, texture, and statistical parameters to provide nuanced evaluation of chilli samples.
5. An automated method for grading chili quality as claimed in claim 1, wherein said flexibility in deployment is achieved by

hosting the classification model on both cloud platforms as well as On-Prem devices.
6. An automated method for grading chili quality as claimed in claim 5, wherein said the chilli grading/classification model hosted on an On-Prem Device (3) comprising of: a) training of model, b) profiling, tuning, compiling and, c) prototyping.
7. An automated method for grading chili quality as claimed in claim 1, wherein said measuring the spectrophotometer color Values samples (8), comprising factors, L*, representing the lightness on a scale of 0 to 100; a* and b*, representing the chromaticity of the samples; and DE*, representing the total color difference value.

8. An automated method for grading chili quality as claimed in claim 1, wherein said method for comprehensively evaluating chilli samples, comprising:
h) applying deep learning methods within a Convolutional Neural Network (CNN) model;
i) the optimized CNN model is deployed on a Neural Compute Stick edge device;
j) extracting features from the chilli samples, including color, texture, and statistical parameters;
k) training the CNN model to recognize and analyze the extracted features;
l) treating a comprehensive dataset for nuanced evaluation of the chilli samples based on the analyzed features;
m) the user can test the query chilli image sample based on chilli inference model on edge device and output the chilli quality;

n) the CNN model’s adaptability allows it to learn and adjust to new chilli varieties, ensuring its widespread effectiveness and versatility.
9. An automated method for grading chili quality as claimed in claim 1, wherein said an edge device for real-time monitoring and analysis, comprising:
d) an optimized convolutional neural network (CNN) model for on-site computation;
e) a Neural Compute Stick (301) integrated on the chip, configured as a neural network accelerator for managing deep learning applications, and;
f) delivering 1 TFLOPS of performance for a small package that fits into hand-held device and delivers high- performance even in severely power-constrained environments.

10. An automated method for grading chili quality as claimed in claim 1, wherein said method involves steps of:
k) an input image, involves a module for efficient collection of chilli sample images;
l) employing wireless sensor-based systems for real-time data acquisition;
m) a pre-processing phase ensuring data quality;

n) normalization & segmentation ensures consistent and standardized input, and enhances the delineation of chilli features respectively, contributing to the accuracy of the grading process;
o) feature extraction methods, encompassing color, texture, and statistical parameters to provide a nuanced evaluation of chilli samples;

p) training of labelled datasets;

q) a database of labelled training images to train the model on a Cluster;
r) make it available as a trained network for remotely operated devices;
s) chilli grading/classification model hosted on laptops, cloud platforms or an On-Prem device (5);
t) a three-grade classification system, distinguishing chilli samples into Chilli Grade 1 (Good), Chilli Grade 2 (Medium) and Chilli Grade 3 (Defect) categories based on the features, encompassing color, texture and statistical parameters.