Suspicious Vehicle Recognition using Number Plate

In most of the surveillance and restricted areas, vehicle recognition has become an essential task due to continuously increasing traffic. It is a very time consuming and tedious task for humans to manually keep a record of all the vehicles entering and leaving a particular area. Each vehicle has its distinctive number plate, which is the combination of special characters. With the help of these number plates, a training-based approach is followed in this paper to detect whether the entering vehicle is suspicious or not in particular area. The dataset of captured images of Indian vehicles is used for training using a deep learning approach. Here, all the images of dataset were trained on the YOLO-v4 object detection algorithm. The algorithm is tested on different vehicle images of different illuminations and variations of light. From the experimental results, this model is able to detect number plate with mean Average Precision (mAP) of 99.6%. Further, the recognition of number plates is done with the help of Optical Character Recognition (OCR) Engine. This system will allow automated entry, restricting suspicious and unwanted vehicles in the restricted areas. With the help of experimental results, we can evaluate the effectiveness of this model.