Study On Detection And Tracking Techniques For Moving Containers In Video Sequences

Currently, computer vision technology is the direction of container recognition and position in automatic container loading and unloading operations. Container-stevedore technology is studied and a simulation system is constructed. In laboratory environment, this paper researches the application of computer vision in container-stevedore and improves the algorithms. All of these algorithms are programmed by C++.Through summing-up and analyzing the characteristics of existing methods in detection and tracking, this paper pays great attention to investigate the practical techniques for moving objects detection and tracking under the condition of a fixed CCD camera, including such techniques as automatic detection and extraction of moving container, tracking of moving containers and so on. The following research has been accomplished:(1) Detection of moving containers in video sequencesIn order to deal with the existing problems in efficiently detection of moving objections, this paper presents a moving container detection and extraction method based on Gaussian mixture model, block-size detection, adaptive threshold segmentation. According to the difference of the consecutive video frames, a real-time dynamic updating scheme of background image or reference image is described. Then, block-size detection is used to get rough area of the foreground. Subsequently, spatial adaptive threshold segmentation is applied to get a more exact foreground. Finally, these two detected results are put together to touch the object of moving container detection.(2) Tracking of moving containerFocusing on the problem caused by time-consuming computation and multi-containers occlusion in object tracking, this paper investigates the basic tracking methods of moving object. The detected foreground is used in back-projection to get the probability distribution map through HSV color model. Subsequently, probability distribution map is used as the input of Camshift algorithm. Additionally, the moving model of a container is established from Kalman filter, which can estimate the position of the container, avoiding the interruption of trajectory if the object is occluded absolutely.In this thesis, video surveillance is investigated and detection & tracking of moving vehicles technology is creatively brought to the container-stevedore. This paper gives a detailed and systematic analysis of the existing mature algorithms in this area through many experiments and successively, constructs the whole algorithm system and achieves the expecting aim. This paper also proposes its own ideas and improvement according to the certain application situation, provides a support to the application system of the automation of container-stevedore.