Research And Implementation Of Road Section Monitoring System Based On YOLCAT Instance Segmentation

With the continuous improvement of people's living standard,the number of cars is constantly increasing,and the number of accidents violating traffic laws and regulations is constantly increasing,which brings great pressure to the traffic control work.Intelligent traffic monitoring system has gradually become an effective way to solve the current complex traffic situation.The main scene of the intelligent traffic monitoring system studied is a common road section,which is a two-way single lane separated by a double solid yellow line;on both sides,there are pedestrians crossing the road and vehicles crossing the line.In view of this complex traffic scene,the research contents of this paper mainly involve:(1)Vehicle shadow removal,this part of the work content is mainly as data enhancement.In this paper,the shadow elimination algorithm of moving vehicle is improved by combining the shadow attributes of gray scale and color map.(2)Vehicle,pedestrian,double yellow line and other targets detection and segmentation.The YOLACT model can be used to detect and segment moving objects in this scene.The experiment shows that the accuracy and robustness of the algorithm model are good.(3)Track the movement of vehicles and pedestrians.Based on the YOLACT segmentation model,the tracking algorithm of moving target trajectory was improved,and the experimental results showed that the tracking effect of this algorithm could reach the expectation.(4)Useful traffic information from the video images is analyzed and output,such as violations by vehicles and pedestrians,and flow statistics of different categories of moving objects in two lanes.A simple improved algorithm based on moving target tracking is proposed,which is proved to be effective and reliable by experimental comparison.(5)Based on the above research results,Microsoft Visual Studio 2016 integrated development environment is adopted as the development platform,and Visual C++ language programming is applied to realize the design of intelligent traffic monitoring system.