The Research On The Detection And Tracking Technology Of The Multiple Motion Objects In The Monitored Area

The technology of detecting and tracking multiple motion objects based on video is the basic and also difficult part in intelligent transportation system. The paper researched the technology in the complicated traffic scene, which was included by different illumination, much shadow and occlusion. The paper included two parts, one was moving objects detection method, and another was the method of tracking motive objects.Parts of moving object detection contained images from color to gray, background initialization, extraction and updating, image segmentation and object recognition, and so on. Images from color to gray used the method of reducing and quantizing the color number. The method made the real color images as the non-real color images with a palette. Then let the non-real color images become gray images. Then the multiple-frame average method was adopted to initialize the background model. Also the existing background updating method was discussed. The traditional background updating method affected by illumination is not accurate. Combined with the adaptive double thresholds and Surendra background updating method, this paper proposed a new Surendra background updating method based on double thresholds. The method also could be called the double threshold motion mask background updating method. It could get good backgrounds, and updated double thresholds adaptively. Meanwhile, it could adapt to varied illumination well.The double thresholds background difference algorithm was adopted based on the background extracted before. Then combined with the mathematical morphology method, the coarse moving object could be obtained. However, the extracted object mixed many shadows. In order to eliminate the shadows well, this paper first introduced several common shadow elimination methods and analyzed their limitations. Then based on the studies, the paper explored an algorithm of removing vehicles’ shadows based on the edge information. The algorithm could eliminate many shadows and better extracted objects in interested areas. Parts of moving objects tracking explored the Kalman technology forecasting the tracking model, and combined with the multiple-feature matching algorithm. The method selected the object area and shape as main features, then used the Kalman filtering technology to forecast tracking model. It could reduce the search scope, and improve the speed of tracking and matching objects. It could not only track objects stably, but also could improve the rate of tracking vehicles in occlusion situations.As a whole, this paper studied the multiple motion objects detection and tracking based on monocular image sequences in traffic scenes. Especially in several aspects, such as the background extraction and updating, shadow elimination and vehicle tracking, they had higher applicable value.