An Intelligent Vehicle Detection Method Based on Tracking Trajectory Analysis

Abstract

The invention discloses a vehicle retrograde intelligent detection method based on tracking trajectory analysis, which uses a primary classifier and a secondary classifier to detect target vehicles in video images, extract target vehicle areas, and assign kernels to target vehicles Correlation filter tracker, each target vehicle is matched with a kernel correlation filter tracker, using the kernel correlation filter tracker to obtain the tracking area of ​​the target vehicle, and then according to the tracking area to obtain the growth direction of the target vehicle's trajectory, and compare it with the initial marked retrograde Directions are compared, if they are the same, the target vehicle is going retrograde; if they are different, the target vehicle is not going retrograde. The present invention realizes the real-time detection of vehicle retrograde through this method, and solves the problems existing in manual detection; at the same time, the reliability of vehicle detection and recognition is improved by using cascaded classifiers, and simultaneous tracking of multiple vehicle targets is realized.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and in particular relates to a vehicle retrograde intelligent detection method based on tracking trajectory analysis. Background technique [0002] With the continuous development of society, there are more and more vehicles on the road, and traffic safety problems are becoming more and more prominent. The number of casualties caused by traffic accidents is shocking every year, and traffic accidents caused by retrograde traffic on the road often cause great harm. Nowadays, the emphasis on road safety is constantly increasing, and a large number of cameras are installed on the road for monitoring. For these massive surveillance video data, in the past, it was mostly manually judged whether there was any abnormality, which would not only consume a lot of manpower but also be impossible 24 Hours of work without interruption. With the development of intelligent video surveillance technology,...

AI Technical Summary

Problems solved by technology

This method is simple to implement, but for the optical flow method, if the slow speed detection of the moving target is a problem, the background modeling is more susceptible to the influence of the environment

At present, in the research methods of vehicle detection and tracking, the traditional methods for detection are based on prior knowledge, using prior knowledge of vehicle bottom shadows, taillights and other prior knowledge to detect vehicles; or using more advanced Haar_like features and HOG features to train classifiers to detect vehicles, but it is difficult to apply a single classifier trained with a single feature to complex and changeable traffic scenarios; or, using the currently popular detection method based on convolutional neural networks, although the effect is significantly improved, but due to the amount of calculation Large and difficult to run in real time on common hardware devices

For the tracking of vehicle targets, most of the previous methods used Kalman filter, particle filter, etc. These methods generally can only track a single target, and once the tracking error cannot be corrected in time, the effect is not reliable

[0004] Therefore, in view of the lack of an efficient and reliable vehicle retrograde detection method in the prior art, the classification reliability of a single classifier in the existing vehicle detection technology needs to be improved, and the existing vehicle tracking method is a defect in tracking a single target, There is an urgent need for a vehicle retrograde detection method based on reliable vehicle detection and tracking technology

Images