tailieunhanh - Effective Method For Detecting Multi Planes From Depth Maps

In the field of visual stereo image processing, plane detection aims to assist in the movement of the mobile vehicle or mobile robot. This paper has carried out the plane detection problem based on depth map by using a new Neighbor Grouping algorithm and a rational Filter. | VNU Journal of Science: Natural Sciences and Technology, Vol. 33, No. 1 (2017) 58-66 Effective Method For Detecting Multi Planes From Depth Maps Dang Khanh Hoa1,*, Bach Ngoc Minh2, Nguyen Tien Dzung1 1 School of Electronics and Telecommunications, Hanoi University or Science and Technology, 1 Dai Co Viet, Hai Ba Trung, Hanoi, Vietnam 2 Vietnam National University, Hanoi, 144 Xuan Thuy, Cau Giay, Hanoi, Vietnam Received 01 November 2016 Revised 15 December 2016; Accepted 23 March 2017 Abstract: In the field of visual stereo image processing, plane detection aims to assist in the movement of the mobile vehicle or mobile robot. This paper has carried out the plane detection problem based on depth map by using a new Neighbor Grouping algorithm and a rational Filter (NGaF). The main advantage of this proposed method is the simplicity while it still ensures the reliability of the results. A concise and clear hypothetical concept of the plane is built in the depth map. Then the plane extracting algorithm is applied based on some strong characteristics of the plane. The results of the applied method are considered positive in terms of both visual assessment and evaluation parameter. Then, the NGaF approach is also be evaluated more superior than the RANSAC algorithm, powerful PPDFDM and FPDIDM methods. The proposed method’s computation time is reduced 33 times compared to the improved RANSAC algorithm (HSBSR). Meanwhile, the result in number of found planes is greater than and PPDFDM, FPDIDM method about 8% percentage. Last, the percentage of calculated valid points is larger than compared methods 2%. It certainly has the ability to implement on common hardware with limited resources as well as to ensure the real-time applications which processes stereo video signal. Keywords: Detection, plane, depth map, depth difference, neighbor point. 1. Introduction disparity map. In [1], the authors combine an improved Hough transform with clustering to search many targets in .