基于深度图像利用BP网络实现遮挡边界检测  

Using BP Network for Occlusion Boundary Detection Based on Depth Image

作　　者：张世辉[1,2] 耿勇 张笑维 ZHANG Shi-hui;GENG Yong;ZHANG Xiao-wei(School of Information Science and Engineering,Yanshan University,Qinhuangdao,Hebei 066004,China;The Key Laboratory for Computer Virtual Technology and System Integration of Hebei Province,Qinhuangdao,Hebei 066004,China)

机构地区：[1]燕山大学信息科学与工程学院,河北秦皇岛066004 [2]河北省计算机虚拟技术与系统集成重点实验室,河北秦皇岛066004

出　　处：《计量学报》2020年第10期1205-1211,共7页Acta Metrologica Sinica

基　　金：国家自然科学基金(61379065);河北省自然科学基金(F2014203119)。

摘　　要：针对视觉目标存在遮挡的现象,提出一种利用机器学习思想检测深度图像中遮挡边界的方法。首先,根据深度图像中像素点的深度信息和空间信息,定义了一种新的遮挡相关特征——最长投影线段特征;其次,设计了一种非线性归一化方法以便对相关特征进行归一化;最后,将遮挡边界检测视为分类问题,利用BP网络对遮挡边界进行检测,并将检测结果进行可视化展示。与其他方法相比,该方法准确性较高,泛化能力较强。Aiming at the occlusion phenomena existing in visual object,an occlusion boundary detection approach is proposed based on machine learning for depth image.Firstly,a novel occlusion related feature named the longest projected line segment is presented according to the depth and spatial information in depth image.Secondly,a nonlinear normalization method is designed to normalize the occlusion related features.Finally,the problem of occlusion boundary detection is taken as a classification problem,meanwhile,the back propagation(BP)neural network is utilized to detect the occlusion boundary and then the detection result is visualized.Compared with existing methods,the proposed approach is more accurate and the generalization performance is better.

关 键 词：计量学 遮挡边界 深度图像 BP网络 最长投影线段 非线性归一化 

分 类 号：TB96[机械工程—光学工程]