A vehicle detection method based on target feature points

Abstract

The invention discloses a vehicle detection method based on target feature points, which belongs to the field of image processing. The invention includes obtaining the transfer matrix between the world coordinate system and the pixel coordinate system; determining the moving target area in the video, extracting feature points in the moving target area to determine foothold points, and selecting stable feature points from the video , track the stable feature points, obtain the trajectory line of the moving object, and judge the trajectory line according to the trajectory line of the moving object. If it is determined that the two trajectory lines belong to the same vehicle, add one to the vehicle counting result, and repeat the above steps until all until the end of the video. By introducing a detection method based on target feature point tracking and clustering in vehicle detection, compared with the existing technology, it reduces environmental restrictions, has high stability and detection accuracy, and reduces detection errors, improving the accuracy of vehicle recognition. Accuracy.

Description

technical field [0001] The invention belongs to the field of image processing, in particular to a vehicle detection method based on target feature points. Background technique [0002] The rapid development of socio-economic activities and urbanization has made both developed and developing countries suffer from huge urban traffic problems. Effective management of existing roads can maximize traffic flow and provide data for road construction. Among them, vehicle detection is carried out on a certain road section, and the traffic flow in a certain period of time is counted, and the traffic flow information of this road section is sent to other end users through Ethernet, which can effectively guide traffic and achieve the purpose of intelligent traffic management. Video-based traffic flow statistics are more and more widely used due to its convenient installation and maintenance and real-time and efficient detection performance. [0003] The currently commonly used video-ba...

AI Technical Summary

Problems solved by technology

However, in practical engineering applications, it is usually necessary to perform vehicle detection and analysis on roads with complex traffic scenes. At this time, due to the serious adhesion between vehicles, the commonly used vehicle detection methods bring large detection errors and reduce the Accuracy of vehicle recognition

Images