DESC:FORM – 2

THE PATENTS ACT, 1970
(39 OF 1970)

COMPLETE SPECIFICATION
(See Section 10; Rule 13)

AN EAR TAG RECOGNITION SYSTEM
FOR CATTLE IDENTIFICATION & METHOD THEREOF

Prompt Equipments Pvt. Ltd
A Company Incorporated under the Indian Companies Act
having address at
501 & 502, Harmony Icon, Hebatpur Road, Thaltej
Ahmedabad, Gujarat, India-380054

The specification in particular describes the nature of the invention and the manner in which it is to be performed:
FIELD OF INVENTION:

The present invention relates to the field of electronic tag recognition for cattle identification.

The present invention provides a tag recognition system for identifying cattle. In particular it provides an ear tag recognition system and method which accurately detects the tag for identification of cattle.

BACKGROUND OF THE INVENTION:

The most prevalent conventional technique for cattle identification in farms is unique identification tags. These tags have a number that helps the farmer to visually identify the cattle. The details pertaining to cattles such as age, breed, sex, height, body color, horn type, tail switch and special marks according to the standards set by the Government can be availed. In the case of milch cattle, it will also have the lactation profile. These tags are generally made from thermoplastic polyurethane elastomer, a material resistant to ultraviolet light and high and low temperatures, and impossible to tamper with once these are sealed. Generally these tags are placed on an ear of the animal and it becomes integral to its body. These tags have been found cost effective.

However, there is no system currently in place to autonomously identify the cattle through ear tag. Alternatively, there exists a system for identification of cattle which involves the use Radio Frequency (RF) transmitters and receivers Instead, the current adopted autonomous identification involves the use Radio Frequency (RF) transmitters and receivers. These RF transmitters and receivers are expensive and make no use of the already in-place ear tag numbers.
Conventionally there are system in place which requires internet facility to function, which is yet not available in all the remotest places. Also in cases, where the internet is available, its continuous availability with requisite speed is a question which cannot be assured; hence the system is not such prevalent and in place.

Yet another system in use is the Optical Character Recognition (OCR) technology; that enables to convert different types of documents including images captured by a digital camera into editable and searchable data. Various systems for OCR tasks include Amazon’s Recognition, Google’s Vision AI, Microsoft Azure’s Computer vision API. These systems are available on the cloud and require an internet connection to work. Furthermore, they are designed for basic OCR tasks and would not give accurate results for the ear tag recognition. Also, these systems requires high end mobiles to compute and provide results hence not user-friendly to a major group of people. This makes it high in cost and also non affordable to many of the users or cattle owners. Moreover, the solution has to be fast enough in terms of computation to applicable in real time.

The end users or cattle owners come from different socio-economic backgrounds. There should be an ear tag recognition system which should cater their needs efficiently and accurately, despite of their socio economic backgrounds. The system should be accurate and at the same time user friendly, cost effective and reliable.

PRIOR ART AND ITS DISADVANTAGES:

An Indian Patent application 201711032865 discloses a method and system for livestock identification wherein a new or existing unique ID is created for an owner and cattle, which is further combined so that ownership of cattle is defined. The present invention is to provide every cattle ID (CID) with complete biometric information, including but not limited to biological traits such as cattle muzzle image, cattle face image, two side profile images, breed, age and gender, wherein all together form a biometric template. In the present invention, unique identifier (UID) of the owner is created which serves the purpose of uniquely identifying an individual such as but not limited to name of the owner and area pin code, Aadhar number etc.

However said system and method at first fail to provide a system that in particular recognizes the existing unique ear tags on cattles. Further the system described herein takes images of the cattle and extracts features to classify and identify cattle. Cattles are highly identical to one another. It is therefore possible that disturbances during the image capture process may occur and cause multiple cattles to look alike. In that case, it is possible to mix up the information of two or more cattles which may lead to errors in final output. Additionally, the system disclosed in the prior requires installation of additional storage devices along with continuous use of internet connectivity and therefore is neither user friendly nor cost effective. Further the response time to provide the output is substantially higher and requires high end user devices and high speed internet connectivity. Further it is prone to errors and recognition requires use of multiple images as only single input does not work.
Another patent application JP2003146066A provides a system for identifying an animal for experiment with which the descent information, the genetic information, and the like, of the animal for the experiment can be managed, and the process information thereafter, and the like, can also be managed. The system for identifying the animal for the experiment comprises attaching a semiconductor chip to the animal for the experiment, registering the peculiar information on the animal for the experiment in combination with the individual ID on a data base, reading the individual ID of the animal for the experiment from the semiconductor chip attached to the animal for the experiment by a reader/writer, and receiving the peculiar information on the animal for the experiment from a server based on the individual ID read by a terminal unit.

However said system involves the use of a semiconductor device to store data which can be read through another hardware receiver. This in turn involves providing additional power supply for the chip and shall incur additional expenses for reader or writer of the same. There are more chance of said system being malfunction and thereby needs frequent maintenance for server and work storage.

Another patent application KR20100005356U relates to an electronic tag necklace for an animal including an IC chip. Unlike a tag mounting device that is directly inserted into an existing living body or attached to an ear, the present invention eliminates the suffering of an animal, thereby removing the animal from suffering. In addition, the present invention relates to an animal electronic tag necklace including an IC chip that prevents stress and removes tissue damage (inflammation) and various side effects (cancer, hair loss, etc.) caused by a wound.

However said tag at first fail to provide a system that in particular recognizes the existing unique ear tags on cattles. It involves the use of a semiconductor device to store data which can be read through another hardware receiver.

Another patent application US2002066418A1 provides the tag includes at least two separate but matching identification means or identifiers which ensure that integrity of the identification is maintained. The electronic identifier is preferably in the form of an electronic transponder which emits a unique identifying signal corresponding to the particular animal. The visual identification means or identifiers may include a bar code, a visual management number, a feed lot number, an electronic identification number and others. Each of these additional forms of identification also match the electronic identifier. The multiple and redundant identifiers provides greater identification system reliability, and further provides users additional options in creating, storing, and manipulating information for a particular tagged animal.

However said prior art relates to combination of two identifying systems one being use of electronic tags similar to afore mentioned prior art and other being visual based identification. Nonetheless sad system relies on barcode scan and reading for visual identification which in many cases can be hindered due to unreadable format of the barcode in general occurrence, thus is not accurate and real-time. Further there is high possibility of mismatch in both tags leading to confusion and lower or no accuracy in the end result. Sometimes it is redundant and hence not efficient.

DISADVANTAGES OF PRIOR ART:
The prior art in public domain either suffers from all or any of the below listed disadvantages:

1. They do not provide a system that effectively utilizes the existing cattles ear tags.
2. Most of these systems require high end electronic devices to work which in turn requires skilled user to operate as well as involves substantial amount of cost.
3. The systems disclosed in the prior art are highly susceptible to provide false positive results and therefore do not provide accurate data and are thus not efficient.
4. Most of the systems disclosed in prior art utilizes various mechanical and electronic components which may include use of transmitters, receivers amongst others which at first requires substantial capital along with skilled user to operate and eventually requires periodic maintenance, in many cases even replacement which in turn is not cost effective and user friendly.
5. Most of the system disclosed in prior art are dependent on continuous internet connectivity in order to function which indeed is troublesome in remotest location where there is no internet connectivity or low connectivity. This dependency on internet connectivity hinders the appropriate utilization of the system and hence is not a collective solution.
6. Even more, many of the prior art systems are unable to work efficiently at low processing power inputs and hence are not effective.
7. Prior art based on internet connectivity, cloud services and high end technical computation are not time efficient.
8. They fail to effectively and accurately identify the cattle details in farm through ear tags with accuracy.

In view of the aforesaid, there requires a system which accurately recognizes an ear tag of cattle and autonomously identifies the cattle in a farm in real time with low cost without internet connectivity which works on low end portable devices.

OBJECTS OF THE INVENTION:

Accordingly the object of the present invention is to provide a tag recognition system for identifying cattle. An aspect of the invention provides an ear tag recognition system and method which accurately detects the tag for identification of cattle.

In an aspect the present invention provides a system that effectively utilizes the existing cattle ear tags and identifies the cattles accurately.

It is the object of the present invention to provide a system for accurate tag identification of the cattles without the need of high end technical components and mechanical components. In an aspect the present invention provides a user friendly system that is easy to operate for the end users.

It is the further object of the present invention to provide an ear tag identification system that accurately detects the cattle data and eliminates any false identification and therefore provide a reliable solution.

In yet another object the present invention provides the ear tag identification system that provides the real time output with substantially low power usage. And is therefore energy efficient and economic.

In the aspect the present invention provides the system that does not require to depend upon the internet connectivity and thus is able to work even in the remotest location with the substantial efficiency.

The object of the invention provides the system to identify the cattles in substantially less time and is user friendly.

In yet another object of the present invention it provides a cost effective ear tag identification system.

In an object the present invention obviates the problems of the prior art and provides an accurate and real time detecting ear tag identification system which saves time, energy and is effective and user-friendly.

BRIEF DESCRIPTION OF DRAWINGS:

Fig.1 : Shows the block diagram of the present an ear tag recognition system
Fig 2a : Shows the sample of Input Raw image data for the present ear tag recognition system
Fig 2b : Shows the cropped ear tag image data obtained from object detection module
Fig 3a : Shows the working of the Recognizing Means of the present an ear tag recognition system
Fig 3b : Shows the working of the Recognizing Means of the present an ear tag recognition system

Meaning of reference numerals:
P : Present ear tag recognition system to identify cattle
1 : Input Module
2 : Tag detection Module
2.1 : Object detection means
3 : Enhancement Module
4 : Localization module
4.1 : Text Localization means
4.2 : Post Processing means
4a : Skew correction means
4b : Binarizing means
4c : Recognizing Means
5 : Classification Module
6 : Display Module

DETAILED DESCRIPTION OF THE INVENTION:

The embodiment of the present invention is to provide an ear tag recognition system for identifying cattle (P). In particular it provides an ear tag recognition system and method which accurately detects the existing ear tags on cattle for their identification. Present invention (P) facilitates the accurate identification of cattle data in real time and eliminates the false results obtained during its identification process. Present invention (P) is not dependent on internet connectivity for its functioning and provides the real time output with substantially low power usage and hence is accurate, energy efficient and user friendly.

Referring to figure 1 to 3b, showing the block diagram of the present invention the invention (P) comprises:
? Input Module (1),
? Tag detection Module (2),
? Enhancement Module (3),
? Localization module (4),
? Text Localization means (4.1),
? Post Processing means (4.2),
? Classification Module (5),
? Display Module (6);

Said Input module (1) connected to the tag detection module (2) is provided to capture the image of the ear tags on the cattle. The input data can be in the form of raw input images. Said Input module (1) may include image capturing means of smart devices (of the user) such as mobile phones, tablets, laptops amongst others. Said tag detection module (2) connected to the input module (1) receives the input from it as raw data images. The tag detection module (2) extracts the relevant information from the ear tag, detecting the exact region of interest to extract the information data of cattle. Said tag detection module (2) comprises an object detection means (2.1) that facilitates to extract the exact data from the raw input images. When the raw input images are transferred to said tag detection module the raw data consists of some noise as shown in Fig 2. Said object detection means (2.1) marks the exact data in bounding boxes and specifies the location of the ear tag in the raw input image facilitating the accurate location of the ear tag as shown in Fig 2a.
Said enhancement module (3) connected to the tag detection module (2) facilitates in enhancing the tag detected and objected by the tag detection module (2). Said enhancement module (3) assists in removal of additional noise form the detected ear tag output obtained from object detection means (2.1). The enhancement module (3) fixes the intensity of the image data input providing clarity of the image data input.

Said localization module (4) connected to enhancement module (3) comprises a text localization means (4.1) and a post processing means (4.2). Said localization module (4) facilitates to extract bounding boxes in regions of the image data input where the text is present. The bounding boxes provided by the localization module (4) is utilized in post processing skew correction. Said localization module (4) recognizes the text in the image data input and localizes the text in said image data input. Said localization module (4) further facilitates in dividing base layer feature map into two halves; one half being a dense block and a transition layer where as other half being integrated with transmitted feature map. Said text localization means (4.1) localizes the text in the image input data with the rotated bounding boxes in diverse orientation.

The localized text data than is allocated to the post processing means (4.2), where skew correction, binarization, removal of false positive values, character box extraction takes place. Said post processing means (4.2) comprises:
? Skew correction means (4a),
? Binarizing means (4b),
? Recognizing Means (4c);

Said post processing means (4.2) at first functions in processing the input data and finalizing it by skew correction means (4a). The text localized at diverse orientation by bounding boxes are bought to the required angle such that it is vertically aligned by said means (4a). Said means than crops the tag from the original image input data providing angle for text boxes ;where the text box with highest score is considered the correct the orientation of tag and is localized there eliminating the false tag orientation. Said Binarizing means (4b) connected to the Skew correction means (4a) facilitates to extract the text bounded in the correct orientation, identifying contours in text localized area. The binarizing means (4b) utilizes the text row boxes to obtain region of interest of the entire row and convert it to gray scale at first and then binarize it to locate the particular text in particular bounding box. Further said recognizing means (4c) filters out the false positive text row boxes with the attributes not limiting to location, area amongst others of the ear tag; providing only the relevant characters to classify further. This ensures the accuracy of the present system. Said recognizing means (4c) recognizes the text in the marked bounded text boxes and provides the string of digits as its output as shown in fig 3a and 3b.

Not to limit the scope of working and implementation of the present invention (P), the working and technical aspect of the said recognizing means (4c) in various aspects to recognize the bounded text boxes is depicted herein below:

? The contours around the bounded text boxes are sorted according to the y-coordinates of their centers and then subsequently sorted into rows hence providing multiple contours in each row of the array. For each such row, if number of contours boxes is very less or very high (with respect to N), they are ignored. Further said means (4c) sorts the boxes according to x-coordinates of center and the median of height, width starting y-coordinate. An array error is kept counting number of errors in each row. Where further each box, the following checks are carried out.
Gaps and intersection between boxes: If the gap or intersection between consecutive boxes is high or low with respect to the median width, errors array is update by one. If gap is of similar size to median width, a box is added in that place. Intersection in real cases happens when one box is below another, in which case one box is deleted and the size of the other is updated as the median height.
Large height: If the height is large compared to median height, error is updated by one and height is changed to median height.
Large width: If the width of a box is large compared to median width (case of mixed characters), the region of interest of the box is extracted from text score map image and the column value where minimum white pixels are located is obtained. The box is divided along the width into two at the column coordinate. If either of the two new boxes has width large compared to median width, it is directly divided into two.
Small size: If the box is very small, it is deleted if the number of boxes in the row is N + 1 or N + 3. If the number of boxes in the row is N or N + 2, the size of the box is increased to median standard.
Height increase: If the height of the box is less than median height but not very less, it is made equal to median height.
If number of errors in a row is more than certain threshold, that row is deleted. If no such pairs are found, the pair of rows with the least number of errors is selected to be returned. If this least error pair has lesser than N-character boxes, the pair with maximum characters is selected. If 1 pair is found which adds up to 2N-character boxes, it is selected as the final pair. If more than one pair is obtained adding up to 2N-character boxes, again the pair with the least errors is selected. Now if the number of errors is equal for these pairs, another value is used. This is the row width. The row width is calculated as the sum of the width of all boxes - the gaps and intersection between them. The row width for an ideal row of numbers must be maximum. Therefore, in the case of pairs having equal errors, the pair with maximum row width is selected. Further each box is added with extra padding on the sides to compensate for morphing. The first box is extended on the left a little more and the last box is extended on the right a little more. If a row contains a certain number of boxes with large area, these boxes can be combined to create a single row. Once the 2N-character boxes have been obtained, each character is cropped and sent to the digit recognizer for classification as shown in Fig 3a and 3b.

Said classification module (5) connected to localization module (4), takes the input form the recognizing means (4c) and resizes and classifies the input image data providing the final output to the display module (6). The resizing is done while maintaining the aspect ratio that in turn prevents the image from getting stretched or compressed along either in way of its dimensions and adding padding if needed to make it suitable to the model input configuration.

WORKING OF INVNETION:

The working of the present ear tag recognition system (P), is herein described in detail with reference to fig 2a 2b 3a 3b. The working involves following steps:
1. At first, when the system is in use said input module (1) of the user smart device, captures the image of the raw input data as shown in Fig 2a.
2. Followed by this the tag detection module (2) crops the raw input data image and extracts the relevant information through object detection means (2.1); as shown in fig.2b. Said object detection means (2.1) marks the exact data in bounding boxes and specifies the location of the ear tag in the raw input image facilitating the accurate location of the ear tag.
3. Further, said cropped image of raw input data is enhanced for visibility and clarity by said enhancement module (3) by removing additional noise. The bounding boxes are than extracted by said the localization module (4) by recognizing the text in image data input and localizes the text in said image data input. It than divides base layer feature map into two halves; one half being a dense block and a transition layer where as other half being integrated with transmitted feature map. The text localization means (4.1) localizes the text in the image input data with the rotated bounding boxes in diverse orientation.
4. The post processing means (4.2) processes said input data and finalizes it by skew correction means (4a) and crops the tag from the original image input data providing angle for text boxes. The text box with highest score is considered the correct the orientation of tag and is localized there eliminating the false tag orientation. The text bounded in the correct orientation, and contours in text localized area is than extracted by the Binarizing means (4b) by utilizing the text row boxes to obtain region of interest of the entire row and convert it to gray scale at first and then binarize it to locate the particular text in particular bounding box. The false positive text boxes are than filtered out by recognizing means (4c) (as shown in fig 3a and 3b) and provides only the relevant characters to classify further ensuring the system accuracy.
5. The input image is further classified by classification module (5) providing the final output to the display module (6).

The present system (P) with its technicalities can operate in online mode as well as offline mode i.e. it eliminates the dependency on the internet connectivity, and continuous internet connectivity and hence can be operated in the remotest location. In case the present system when operated in online mode said detection module (2), enhancement module (3), localization module (4) and classification module (5) are processed and functioned on server. Said input module (1) functions through Application programming interface (API) providing output to display module (6). The present system when operated in offline mode all the modules of the system are processes and operated on the user device through optimization such that unnecessary operations are eliminated to avoid repetitions, helps in substantially fast processing. ,CLAIMS:WE Claim;

1. An ear tag recognition system (P), wherein said system (P) comprises;
An Input Module (1),
A Tag detection Module (2),
An Enhancement Module (3),
A Localization module (4),
A Text Localization means (4.1),
A Post Processing means (4.2),
A Classification Module (5),
A Display Module (6);

Said Input Module (1) is configured to capture and provide the raw image data input of the ear tags on the cattle; said Tag detection Module (2) connected to input module (1) is configured to extract the information of ear tag from said raw image data input, detect the exact region for information through an object detection means (2.1); Said enhancement module (3) connected to the tag detection module (2) is configured to enhance the tag detected and objected by the tag detection module (2), remove noise and fix intensity of image; Said localization module (4) connected to enhancement module (3) is configured to extract bounding boxes in regions of the image data input where the text is present ; Said classification module (5) connected to localization module (4) is configured to take input from localization module (4) resize and classify and provide final output to display module (6).

2. The ear tag recognition system (P) as claimed in claim 1, wherein said Tag detection Module (2) comprises an object detection means (2.1);configured to eliminate the noise from the raw data and extract data from the raw data input image.

3. The ear tag recognition system (P) as claimed in claim 1, wherein said localization module (4) comprises a text localization means (4.1) and a post processing means (4.2).

4. The ear tag recognition system (P) as claimed in claim 1 and claim 3, wherein said text localization means (4.1) is configured to localize the text in the image input data with the rotated bounding boxes.

5. The ear tag recognition system (P) as claimed in claim 1 and claim 3, wherein said post processing means (4.2) configured to process and finalize input data through skew correction comprises:
A Skew correction means (4a),
A Binarizing means (4b),
A Recognizing Means (4c);

Said Skew correction means (4a) is configured to localize text in bounding boxes, crop the tag from the original image input data providing angle for text boxes; Said Binarizing means (4b) connected to the Skew correction means (4a) is configured to extract the text bounded in the correct orientation, and identify contours in text localized area convert it to gray scale and binarize text; said recognizing means (4c) is configured to filter out false positive text row boxes.

Dated this 31st Day of May 2022

Gopi Trivedi
IN/PA 993
Authorized Agent of applicant

To,
The Controller of Patents,
The Patent Office
At Mumbai