Vehicle occlusion treating method based on Markov random field

Abstract

The invention discloses a vehicle occlusion treating method based on the Markov random field. The method comprises step 10), establishing a background image; step 20), establishing a vehicle foreground profile; step 30), filling blocks in a vehicle outline; and step 40), performing occlusion treatment on the blocks, including step 401), in all the n blocks contained in a vehicle K, taking the i block in the (t-1) frame gray scale image as a template, and determining a the most matching block in the t frame gray scale image by using a three-step search method; step 402), measuring motion vectors; step 403), performing verification treatment on the matching blocks; step 404), performing occlusion judgment and occlusion treatment on the blocks; and step 405), returning to the step 401) until i=n to finish the verification treatment and occlusion treatment of all the blocks. The vehicle occlusion treating method is high in vehicle tracking accuracy, accurate in treatment and good in timeliness.

Description

technical field [0001] The invention belongs to the fields of video image processing, computer vision and intelligent traffic monitoring, and specifically relates to a vehicle occlusion processing method based on Markov (English translation: Markov) random field. Background technique [0002] With the continuous development of economy and society, the number of motor vehicles in our country is increasing rapidly, traffic problems are becoming more and more prominent, and traffic accidents occur frequently. The role of traffic monitoring is particularly important at the moment. With the continuous development of science and technology, traffic monitoring has changed from the traditional over-reliance on people and hardware equipment to the direction of intelligence and informatization. Vehicle detection based on video technology And tracking has become a hot spot in the field of intelligent traffic monitoring. [0003] Compared with the traditional vehicle traffic informatio...

AI Technical Summary

Problems solved by technology

Of course, it also has some limitations. When multiple vehicles appear in the field of view at the same time, occlusion between vehicles is easy to occur, and occlusion will greatly reduce the accuracy of vehicle tracking.

How to improve the accuracy of vehicle tracking has become a research hotspot in this field. At present, there are mainly tracking methods based on feature points, tracking methods based on contours, tracking methods based on 3D models, and tracking methods based on optical flow. These methods have Certain limitations, low tracking accuracy, and the actual application effect of the above methods in vehicle tracking still needs to be improved

Images