Based On The Infrared Video Sequences Of Pedestrian Detection And Tracking Method Research

Pedestrian detection and tracking technology is not only of great significance in traffic control, traffic monitoring and intelligent vehicle auxiliary system,but also can be widely used in the field of video surveillance such as banks,supermarkets, museums, railway stations, car parks and restricted military zones. Target environment can be monitored by infrared camera. With its penetration ability,the result of all-weather, long-distance observation and less affected by smoke, fog, rain, snow can be achieved. Therefore, the research of infrared pedestrian detection and tracking has been carried out in this thesis is not only has important theoretical and practical meaning, but also has potential economic value and broad application prospect.Based on pedestrians in infrared video sequences, the research of infrared pedestrian detection and tracking has been carried out in this thesis. The main contributions of this thesis are as follows:(1) The basic knowledge of the infrared image preprocessing has been introduced. Through the analysis of the characteristics of the infrared image and its noise, the infrared image filtering technology, and an overview of commonly used methods of image segmentation have been discussed in this thesis.(2) The traditional infrared pedestrian detection methods have been outlined. Considering the background subtraction algorithm can’t be adaptive when background changes, while the temporal differencing algorithm will get some pseudo-targets or empty holes when pedestrian target moving too fast or backgroud changes of high frequency, these two methods were combined to differential infrared images. Then iterative threshold algorithm was uesd to segment the difference images. At last, based on the simple feature discrimination method, detection and identification of the target area has been processed.(3) The traditional infrared pedestrian tracking methods have been outlined. Based on pedestrian detection and identification, the traditional Paticle Filter tracking method has been improved in the selection of the target feature in this thesis. It weighted gray feature information with spatial information and extracts target motion characteristics through image differnence.A state observer model was established by these two fearures.To make the most of these two features, they have been fused together in the classical model of Particle Filter in this thesis. In order to make sampling particle converges to the real state of the target area, the Mean Shift theory was embedded in the classical model of Particle Filter.(4) Multiple sets of simulation have been carried out, such as filtering denoising, differential detection, thresholding segmentation, target identification and pedestrian tracking. The experimental results show that the improved algorithms adopted in this thesis did well in pedestrian detection and tracking targets under a variety of conditions.In summary, infrared pedestrian detection and tracking problems are researched in this thesis. The experimental results show that the improved algorithms adopted in this thesis did well in pedestrian detection and tracking.