The Research On Next Best View Based On Self-occlusion Information In Depth Images

Self-occlusion is caused by one part of the object itself occluded the other part of the object,which can be seen everywhere in daily life.With the deepening of research on computer vision,and the popularity of application on depth image,the problem of next best view based on self-occlusion information in depth image is becoming a hot topic.On the basis of the comprehensive analysis of domestic and foreign research status,the next best view based on self-occlusion information in depth image has been deeply discussed in this paper.Firstly,for the visual object is stationary,a next best view method based on self-occlusion information of visual object is proposed.In this method,the self-occlusion detection result is utilized to model the set of occlusion line segments,and the best observation position of each occlusion line segment is calculated and all best observation positions form a set.Then,a minimum spanning tree is build up based on the set of best observation position and the longest path in the minimum spanning tree is gradually cut,and the set of candidate best views is formed by using mean shift algorithm in each tree.Finally,the next best view is determined by maximizing the objective function.Secondly,for the visual object is moving,a next best view method based on self-occlusion information of visual object is proposed.Firstly,a depth image of moving object is acquired and the self-occlusion detection is utilized in the acquired image.On this basis,the self-occlusion regions are modeled by utilizing space quadrilateral subdivision.Secondly,according to the modeling result,a method based on the idea of mean shift is proposed to calculate the result of self-occlusion avoidance corresponding to current object.Thirdly,the second depth image of moving object is acquired,then the feature points in two images are detected and matched,and the 3D motion estimation is accomplished by the 3D coordinates of feature points which are matched.Finally,the next best view is calculated by combining the result of self-occlusion avoidance and 3D motion estimation.At last,the validation of feasibility and effectiveness are performed by experiments,and the comparison and analysis of the experiment results are present in the paper.