Research On Key Technology Of Container Detection And Location Based On Computer Vision

To achieve terminal automatic,intelligent container terminal is the future development trend of shipping.However,there is still a lack of stable and efficient way for loading and unloading of bridge crane.Therefore,the paper focus on the recognition and positioning of container in automatic port stevedoring and it puts forward the corresponding technical methods based on the container corner piece positioning and tracking detection,container corner pieces accurate positioning and container positioning model.In this paper,the main work and achievements are as follows:First,we propose a rough detection method of container corner pieces based on HOG feature and SVM classifier to solve the problem of corner piece rough positioning.The algorithm gives some strategies in the aspects of sample acquisition,image feature selection,classifier training and multi-scale sliding window detection and achieves about 95% detection accurary in the different types of container images,even in the light interference and shadow conditions can get better detection results.And it give the corresponding solutions in the loss of detection and tracking of corner piece.Second,the purpose of the accurate positioning of the corner piece based on iterative fitting is to obtain the intersection point of the straight line of the two edges of corner piece.And the inflection point is replaced by intersection point to achieve accurate positioning of the corner piece.The first step is to extend the corner piece images.Secondly,the image can be segmented by setting the known foreground,background area in advance.Then,the partial edge point set of the corner piece is obtained by setting the extraction area.Finally,using the iterative fitting to remove the error point and achieving accurate straigt line fitting results and getting the intersection coordinates.Third,the main purpose of container model and error modeling is to correct the error in image positioning algorithm and binocular stereo vision system.In order to fully calculate the deviation which generated in the positioning process,each step in the whole process is systematically analyzed.In the image positioning algorithm,the most error is generated in image segmentation and line fitting and this paper carries on the modeling analysis.In the binocular stereo vision system,the error arises in the hardware manufacture,the installation and distance,angle deviation among each plane.Finally,achieving the accurate positioning of container by correcting error which is generated in image positioning algorithm and binocular stereo vision system.