Research And Application Of Target Localization Algorithm Based On Binocular Vision

In the measurement tasks based on binocular vision,the problem is mostly solved by measuring relative information.How to achieve measurement of the absolute localization of the target has become the focus of this thesis.In binocular vision system,the stereo matching algorithm is one of the important factors affecting the accuracy and efficiency.In the face of complex scenes,the accuracy of stereo matching algorithm is poor,and in the face of images with large depth range,its matching efficiency is low.In automated measurement,the software algorithm needs to select representative points within the pixel range of the target for measurement.When there is occlusion between targets,the selection of representative points is often wrong.Aiming at the above problems,this thesis studies the target absolute localization method based on binocular vision.The main research contents and work of the thesis include the following three aspects:1.Aiming at the application of stereo matching algorithm in outdoor conditions,a semi-global stereo matching algorithm with matching cost fusion and hierarchical iterative optimization is proposed.First,the algorithm adopts a hierarchical iterative matching strategy as a whole to optimize the search space.Second,using the fusion cost calculation method makes the initial cost more accurate.Finally,the strategy of multi-core parallelism and data parallelism is used to give full play to the hardware performance to accelerate the calculation process.The experimental results show that the matching accuracy of the method and the computational efficiency on the CPU platform are improved.2.Aiming at the absolute localization problem of target,a target segmentation and absolute localization algorithm based on binocular stereo vision is proposed.Firstly,combining the target detection algorithm with the depth information,an improved target segmentation method based on visibility is proposed,and the segmentation results of all the targets in the scene are obtained by this method.For each target,select representative points in its segmentation results to calculate its camera coordinates.Finally,the coordinates of the points are collected by the real-time kinematic localization device to establish the conversion relationship between the camera coordinate system and the geodetic coordinate system,and the absolute localization of the target is completed.The experimental results show that the proposed target segmentation method can effectively segment independent and occluded targets,and the absolute localization method also has a higher localization accuracy.3.Considering the errors in practice,the selection principle of common points in the calculation of geodetic coordinates is studied.In this thesis,real-time kinematic localization equipment and self-built binocular localization system are used to collect multiple sets of common point coordinates for experiments.According to the absolute localization method used in this thesis,the influence of different number of common points on binocular localization is compared and analyzed.The experimental results show that in engineering practice,the coordinate transformation using more common points can ensure the stable accuracy of the localization system.