Vehicle Counting and Detection

Intelligent vehicle identification and computing has become an important factor in highway management. However, due to the difference between vehicles, their search remains a problem that directly affects the accuracy of vehicle counting. To solve this problem, this article presents an image-based vehicle tracking and counting system. The study published new information about highways with 57,290 descriptions in 11,129 images. Compared to the available publicly available data, the proposed data has little in-image description, providing complete data for in-depth investigation of vehicle detection.In the proposed vehicle tracking and counting system, first the road pavement in the image is removed and a new proposed road segmentation is divided into far and near areas; This method is important for the development of car detection. Then place the above two regions in the YOLOv3 network to determine the type and location of the vehicle. Finally, the vehicle trajectory is obtained with the ORB algorithm, which can be used to determine the driving direction of the vehicle and to obtain different vehicles. Several main video analysis methods are used under different conditions to analyze the proposed process. Experimental results prove that the use of the proposed segmentation method can provide more detection, especially for the detection of small objects.Also, the new concept explained in this article is very effective in travel and traffic decision. This form is very important for the control of detail scenes..