Research On 3D Segmentation Of Indoor Objects Based On RealSense

Image segmentation is the process of dividing an image into several specific regions according to certain rules,and then extracting these regions.With the rapid development of computer graphics and 2D and 3D image technology,image segmentation technology has been widely used in various fields as one of the important basic steps.However,most of the current image segmentation work is mainly focused on the processing of two-dimensional images.As a device that can acquire both color and depth images,RealSense stereo depth cameras provide depth information in the depth images for the current research on 3D images.The use of RealSense for research has become a trend.To this end,based on the GrabCut image segmentation algorithm,this paper combines the depth information obtained by the RealSense camera on the basis of the two-dimensional image segmentation algorithm to study the three-dimensional image segmentation technology.The main work is as follows:First,in order to improve the segmentation efficiency and segmentation effect of the GrabCut algorithm when the picture is rich in details,in the preprocessing process,the SLIC algorithm(linear iterative clustering algorithm)is used to introduce superpixels instead of pixel estimation parameters of the Gaussian mixture model to construct a network flow Figures:Use popular sorting algorithms to perform background prior and foreground prior to generate saliency maps,fuse and optimize the saliency maps,and obtain threshold segmentation results by thresholding the saliency maps to make the foreground and background more separated.The SLIC algorithm can effectively improve the algorithm segmentation efficiency and reduce the segmentation time.Then,this paper proposes a GrabCut algorithm that merges depth information,transforming the ordinary two-dimensional segmentation algorithm into a three-dimensional segmentation algorithm.The depth information is extracted from the depth image obtained by the RealSense depth camera,and the saliency map is guided by the depth information to initialize the algorithm;then the depth information is introduced into the energy formula of the GrabCut algorithm.Depth information can describe the distance between the object and the source of the photograph,and more accurately distinguish the target object from the background.The improved algorithm is compared with other image segmentation algorithms.Experiments show that the use of depth information and color information can make the segmentation results more accurate.The new three-dimensional image segmentation method proposed in this paper effectively improves the segmentation results,improves the segmentation efficiency and accuracy of the algorithm when the background and target objects have similar color differences or there are shadows in the figure.Provides a new way of thinking.