Occlusion Detection And Avoidance Methods Based On Depth Image

Occlusion phenomenon can be seen everywhere in human’s daily life.When visual objects along the observer’s observation directionbecame staggeredin spatial structure,it will inevitably lead to the disturbance of observers observed behavior.For example, in the pattern recognition, automatic scene recognition, three-dimensional reconstruction and other applications in the scene, the occlusion phenomenon will bring great disturbance.In recent years, the deep image which contains the characteristics of the three-dimensional information of the observation object make it rapidly developed and applied,which? makes how to solve the urgent problem based on depth image detection and occlusionavoidance.Based on the comprehensive analysis of domestic and foreign research status, this paperdeeply research the occlusion detection based on depth image and the way of calculating the next best observation.Firstly, this paper introduces the definition and collection method of depth image, the definition and generation of occlusion;meanwhile, it also introduces the idea of the common method of occlusion detection, and the concept of occlusion avoidance;then the paper introduces the principle of the classifier based on the idea of machine learning in occlusion detection methods, and the basic principle of the gradient descent method.Secondly, in the process of researchingocclusion detection methods of deep image, diggingdeeply the objects’ form in 3D space presenting the morphologyand the three-dimensional coordinates of the distribution when it occurs occlusion phenomenon.It can completely find a kind of occlusion boundaries of occlusion phenomenon.Thirdly, based on the detection results of occlusionboundary, establishing the model of external surface of occluded region to describe the occluded regions unknown information,constructinga math model of the next best view,and using gradient descent optimization algorithmto solve themath model of the next best viewto achieve the occlusionavoidance.Finally,the feasibility and effectiveness of the proposed occlusion boundary detection and avoidance method are verified by experiments and compared with other methods, the experimental results are analyzed.