Segmentation Of Space Scene Based On 3D Camera

With the development of science and technology and the development of robot more intelligent, and the pursuit of the robot's autonomy and intelligence is currently in the field of robotics research focus. While the robot in an unknown environment execution autonomy and intelligence tasks, the need of environment need to be understood and need to like the human brain as r ecognition and understanding existing in the current field of vision or environmental objects and object, perception of their environment. Therefore, intelligent mobile robot perception and recognition ability of the environment in which to become importan t research direction of the robot autonomy and intelligence development. Image segmentation is a key technology for robot with perception and recognition ability to the environment. With the rapid development of 3D laser ranging technology, there have been many capable of acquiring 3D information in the 3D scene of a 3D camera, before for two-dimensional image segmentation method is difficult to apply to the 3D information of image segmentation based on, therefore, more researchers gradually into based on image processing techniques for 3D spatial information research. Due to the 3D camera gradually applied to intelligent mobile robot field. Therefore, the study of the a way to make intelligent mobile robot can perceive the objects in the environment or to identify view scene object three-dimensional segmentation technique is very necessary.The plane and the surface is the main part of the 3D scene. In the artificial scene mostly flat. Research on 3D scene plane and curved surface fitting to the objects in the environment out of spiritual domain segmentation and recognition, perception and recognition and can be applied to robot. This paper studies the characteristics of 3D scene, the plane segmentation based on three-dimensional information of 3D scene, and with the 3D plane area and combined with the merging process optimization of fuzzy rules.Firstly, according to the factors of Kinect depth imaging principle and the hardware and software of 3D camera, analyze some problems existing in the 3D data acquired by Kinect 3D camera, and puts forward the three-dimensional data denoising smoothing and repairing method for repair of black hole depth image bilateral filtering method. Secondly, there are many characteristics of plane in 3D space based on a space plane detection method based on three-dimensional information, the method based on the 3D spatial information acquired by 3D camera, based on the normal vector, vector retargeting rule to technology and the plane area estimation based on local surface fitting method, detection and recognition of 3D planar region, aiming at the over segmentation problem of planar segmentation, and the segmentation of plane combined treatment. Finally, in view of the existing space description of 3D surface when the surface is irr egular, or the objects exist and in 3D space, the B spline surface fitting algorithm. In order to improve the accuracy of 3D segmentation, presents a method of curved surface fitting based on fuzzy decision rules, the method of a fuzzy decision rules, to d etermine the need for plane fitting surfaces using the rules based on the area, and use B- spline surface fitting algorithm for the fitting. The method realizes the over segmentation problem of the optimization of 3D segmentation plane detection based on the appearance, and to solve the plane of space curved surface and irregular objects can 't be a complete description of the problem.In order to achieve the proposed segmentation technique for space camera based on 3D visualization, 3D segmentation and imp rove the process of human-computer interaction, the design and implementation of a software system of 3D scene segmentation based on Kinect space camera, the software system includes preprocessing of 3D spatial data based on 3D plane detection, and based o n the three-dimensional information the plane of the surface fitting technology, can realize the data acquisition mode, set the camera to view the Kinect step by step real time 3D segmentation and 3D real-time preservation of the segmentation results and o ther functions.