Research On Vehicle Detection And Tracking Method Based On Computer Vision

With high practical application value,both vehicle detection and tracking technology can be applied into many fields,such as: traffic scenes,automatic driving,intelligent video surveillance,and so on.These two technologies belong to a sub-category of moving target detection and tracking in the field of computer vision.However,due to the complexity of vehicle operation environment,many disturbances around the vehicle and unstable running speed,vehicle detection and tracking are still facing many challenges.which makes the detection and tracking of vehicles still face many challenges.Stable tracking of vehicles is an urgent problem to be solved.In view of this,the main works of this paper are:In the aspect of moving vehicle detection,the mainstream target detection algorithm is analyzed,and the Vibe algorithm with fast running speed and simple algorithm principle is selected to solve the problem of moving vehicle detection.Aiming at the problem that the detection result of Vibe algorithm is easy to introduce ghost region,which can not judge the shadow region,and the detection error is large under dynamic background,this paper proposes an improved scheme.Combining the three-frame difference methods,it fills out the true and complete background image,establishing a background model for each pixel,and quickly eliminating the ghost region;under the C1C2C3 color space model,the detection result of the Vibe algorithm is judged twice,effectively eliminating the background area in the detection result;those pixels are classified by using the dynamic threshold,and the threshold can be adjusted by the background dynamic change,which reduces the influence of the dynamic background on the detection result.Compared with the original Vibe algorithm and other target detection algorithms,the experimental result verifies that the improved algorithm suppresses ghost regions and eliminates shadow regions in fewer frames,improving the overall performance of moving vehicle detection.In terms of vehicle tracking,this paper makes full use of the advantages of neural network,and proposes a target-scale adaptive tracking algorithm for converged symmetric networks to achieve stable tracking of vehicles,and designing a convolution symmetric neural network to achieve target center positioning,then the target tracking problem is transformed into the problem of measuring the similarity between the template image and the candidate region,and the network is fine-tuned online in combination with the dynamic update and static update strategies,Then,the target center is multi-scale sampled,as well as the most suitable scale is selected as the tracking window.Finally,the tracking performance of the algorithm is tested on the OTB public dataset,and compared with the other nine mainstream tracking algorithms,the algorithmcan achieve stable tracking of vehicles under various interference factors.