Description:TRAFFIC SIGN RECOGNITION AND VOICE ALERT SYSTEM USING CNN
Field of Invention
The present invention is a IOT field. The project is based on deep neural network model that can classify traffic signs present in the image into different categories.
The Objectives of this Invention
To develop an efficient and effective model, which predicts the traffic signs boards with best accuracy by using raspberry pi technology and CNN model to classify the image categories and to display the traffic sign board to give voice alert with accuracy.
Background of the Invention
In (US2019/10816993B1), a three-dimensional model is created using the results of the camera and sensors, and a vehicle is then used to travel the road using the model that has been created. This is one of the methods used by smart cars for navigating a road. In another invention (US2020/11640174B2), mechanism used in smart cars for self-navigating that involves building a 3D model using camera and sensor outcomes, retrieving a collection of high-definition maps, and producing a trip containing sections that travel from point A to point B; utilizing a camera and a sensor, identify a motorway entry or exit lane according to a road marking; if the route segments reaches the point of entry or exit, stay in the present lane before quitting; alternatively, proceed towards the entry or departure.
In (US2021/0349460A1), The automobile's sensor(s) may provide information from the sensors, which may then be encrypted to produce encrypted sensor information. To show on the control system's augmented reality headset, the digitally encoded data from sensors may be sent to the control system. Controlled inputs via the control unit may be represented in the data for control that the engine receives from the control panel, and the control signals can trigger one or more of the vehicle's actuator elements to be activated. In additionally (DE2019/112019006468T5), A wide variety of depth forecasting instruments, including but not limited to RADAR sensors, LIDAR sensors, and/or SONAR sensors, can be used to create and encode data that can be used to train the DNN. To reconfigure the DNN for usage with image information obtained from cameras with changing criteria, such as B. changing fields of view, several instances of camera adaptability techniques might be implemented.
In (Choda et al [2023], 2023 International Conference on Intelligent Data Communication Technologies and Internet of Things (IDCIoT), Bengaluru, India, 2023, pp. 445-450), This article outlines a method for a traffic sign recognition system. The many machine learning algorithms that are very accurate in predicting and recognizing real-time roadway signs have been critically examined in this research. It is concerning how frequently fatalities occur on our nation's streets. When someone ignores traffic signs and warnings, they endanger both their own life and the lives of passengers by (Visaria et al [2022], 2022 2nd International Conference on Intelligent Technologies (CONIT), Hubli, India, 2022, pp. 1-6).
Summary of the Invention
The idea for the system was developed and tested with voice alert and traffic sign board recognition. The proposed approach makes use of a hardware component that the user can manipulate to identify the traffic sign (as an animated model for a vehicle). The suggested solution utilises the raspberry pi camera, that can be employed as a standard camera, to take real-time pictures.

Detailed Description of the Invention
We have come up with a model which uses Raspberry pi technology in CNN model for image processing and traffic sign recognition The goal of the model is to capture the traffic signs with the help of IoT devices and send the captured image for image processing. We use a CNN model which explores the dataset which is provided. Then we train and validate the model before testing it with the actual input. The captured image is now processed and analyzed using the CNN model to give the result as an alert on the user’s screen.
The raspberry pi camera is used to capture the traffic signs and the captured images are sent to the image processing algorithm. The Algorithm process the image and displays the results. The algorithm gives alert by displaying the traffic sign details on the user’s screen. We use image sensors to collect images of the traffic signs. The data from the sensor is converted to readable form using raspberry pi technology which is an open-source technology. The sensors used are analog sensors. For some of these analog sensors, we use special packages in raspberry pi to establish connectivity. Using the raspberry pi, we can send or process images. The images can be stored in a file for later use or can be retrieved immediately. Once the raw images are retrieved, they are analysed using image processing algorithm. We will analyze the images of traffic signs in different categories and analyze the results using machine learning algorithms to draw a conclusion about the recognized traffic sign.
The first step is to capture an image of a traffic sign using a camera. This image will be the input for the CNN. Before the image can be fed into the CNN, it needs to be pre-processed to make it suitable for the network. This may involve resizing the image, normalizing the pixel values, and converting it to grayscale or color. The pre-processed image is then fed into the CNN for a forward pass. The CNN is a type of neural network that is designed to process images and extract features from them. Feature extraction: The CNN extracts feature from the input image by applying a series of convolutional filters to the image. These filters identify patterns and edges in the image, and build a hierarchy of features that represent different levels of abstraction. Once the features have been extracted, the CNN uses them to classify the input image into one of several possible traffic sign classes. This is done using a series of fully connected layers at the end of the CNN. Finally, the CNN produces an output that indicates the predicted traffic sign class for the input image. This output can be displayed on a screen or used to control a vehicle or traffic signal.
Traffic Sign Recognition (TSR) and Voice Alert System using Convolutional Neural Networks (CNN) typically follows the following working model with a real time usecases:
Dataset Preparation: A large dataset of traffic sign images is collected and labeled. The dataset should cover various types of traffic signs, different lighting conditions, and weather conditions to ensure robustness of the model. Data Preprocessing: The collected dataset is preprocessed to enhance the quality of the images. This may involve resizing the images to a standardized resolution, normalizing the pixel values, and applying techniques such as histogram equalization or color balancing to improve image clarity.
Model Training: A CNN model is constructed and trained using the preprocessed dataset. The CNN architecture typically consists of convolutional layers, pooling layers, and fully connected layers. The convolutional layers learn spatial features of the input images, while the fully connected layers provide classification based on these features. Training the CNN: The prepared dataset is divided into training and validation sets. The CNN model is trained on the training set using techniques like stochastic gradient descent (SGD) or Adam optimization. The model's weights are adjusted during the training process to minimize the difference between predicted and actual labels. Evaluation: The trained CNN model is evaluated using the validation set to assess its performance. Metrics such as accuracy, precision, recall, and F1-score are calculated to measure the model's effectiveness in recognizing traffic signs. Testing and Deployment: The trained CNN model is tested on a separate testing dataset to assess its real-world performance. Real-time traffic sign images are fed into the model, which predicts the type of traffic sign based on learned features. Voice Alert System Integration: Once the traffic sign is recognized, a voice alert system is integrated into the setup. The system can be implemented using text-to-speech technology, where the recognized traffic sign is converted into an appropriate voice message. The message is then communicated to the driver or relevant authorities through speakers or audio devices.
Real-time Implementation: The entire system, including the trained CNN model and the voice alert system, is deployed in a real-time scenario. The system continuously captures video frames or images from a camera, processes them using the CNN model for traffic sign recognition, and generates voice alerts when a traffic sign is detected. It's important to note that this is a simplified overview of the working model, and there may be additional steps or optimizations involved depending on the specific implementation and requirements of the traffic sign recognition and voice alert system.
The Traffic Sign Recognition (TSR) and Voice Alert System using Convolutional Neural Networks (CNN) offers several key functionalities: Traffic Sign Recognition: The primary functionality of the system is to accurately detect and recognize traffic signs from real-time images or video streams. The trained CNN model analyzes the input images and classifies them into specific traffic sign categories based on the learned features. This functionality helps drivers and autonomous vehicles to identify and understand traffic signs present on the road. Real-time Performance: The system operates in real-time, meaning it can process incoming video frames or images on the fly and provide immediate recognition results. This enables quick responses to changing traffic conditions and ensures timely alerts and notifications. Voice Alert Generation: Once a traffic sign is recognized, the system generates voice alerts to provide timely information to the driver or relevant authorities. The voice alert system converts the recognized traffic sign into an appropriate voice message, which is then communicated through speakers or audio devices. This functionality ensures that the driver is informed about the detected traffic sign, helping to enhance road safety and compliance. Overall, the Traffic Sign Recognition and Voice Alert System using CNN combines the power of deep learning and voice technology to accurately recognize traffic signs in real-time and provide voice alerts, contributing to improved road safety and driver awareness.

4 Claims & 1 Figure
Brief description of Drawing
In the figure which is illustrate CNN model image processing of the traffic sign recognition.
Figure 1, Process of extracting the traffic sign details using CNN. , Claims:The scope of the invention is defined by the following claims:

Claim:
1. A system/method for predicting the traffic sign board by recognizing the image, said system/method comprising the steps of:
a) The system starts up, stream of data is collected by the camera (1). The micro controller (2) is then provided with the information it has collected feed. It then stores the captured images (3) into the system.
b) The CNN method is used for incorporating a novel model that has been trained into the proposed innovation with different images (4), then the prediction can be done by proposed innovation (5).

2. As mentioned in claim 1, the various traffic signed image is captured by the micro control enabled camera, that will be stored into the system.
3. According to claim 1, the user needs to handle the raspberry pi OS for processing the image and to display the result and to give the voice alert to the user.
4. As mentioned in claim 1, the stored images will be given to the CNN model to predict the sign and announce via voice control system.