Automatic Detection for Moving Car on the Road Using YOLOV5 Algorithm

Abstract

A camera vision system is a technology that utilizes image processing and video surveillance techniques to gather real-time and video input data. This technology is capable of detecting and classifying objects in real-time, whether through taking pictures or recording videos. The older versions of YOLO, such as YOLOv2 and YOLOv3, may have limitations that can lead to errors in data collection. These errors can cause inconvenience for both data collectors and road users, which can negatively impact the reputation of the system. This work aims to create a camera vision system that incorporates advanced algorithms to improve the accuracy of object detection and tracking. Specifically, the system utilizes the algorithm of YOLOv5 and Deep SORT to detect and track various classes of objects with a high degree of accuracy, thus reducing the risk of errors in data collection. This will enhance the performance of the system, and also make it more reliable for its intended purpose. The primary objective of this work is to develop a robust and efficient camera vision system that can be utilized in a wide range of applications. The system will be designed to improve safety and efficiency and will be suitable for tasks such as traffic monitoring, by providing accurate and real-time information about the traffic situation. For object detection on YOLOv5, a high accuracy of at least 90% is ideal, along with a low false positive rate. The model should be able to detect and classify objects in real-time with minimal latency. It should be able to recognize a diverse range of objects and perform well under various conditions, such as different lighting, camera angles, and object sizes. Additionally, the model should be able to handle occlusions and multiple objects in the same frame with high precision, providing accurate object boundary boxes and confidence scores.