Image segmentation and object measurement based on stereo vision

Stereo vision is an effective measurement method, which uses the principle of triangulation to obtain the three- dimensional information of the surface points of the object. Image pre-processing is the processing of input image before segmentation and matching in image analysis [1], which the core issue is Segmentation. Only accurate segmentation of the measured object can accurately measure its volume. Firstly, the measurement volume generates three-dimensional point cloud; Secondly, the object contour was obtained and traversed from the point cloud, and the largest and smallest contour was filtered. Secondly, the smallest rectangle of the contour and the coordinates of four vertices of the rectangle are obtained. Finally the size of the object was calculated. Only volume of the object was accurately measured, it can provide a more efficient basis for the grasping strategy of the manipulator. In this paper, the accuracy of the volume measurement of the object is obviously improved by the improved vision algorithm.