Algorithm and Experiment of Cocoon Segmentation and Location Based on Color and Area Feature

Abstract

To improve the lower efficiency of silkworm cocoon harvesting, an algorithm of cocoon image segmentation and coordinate location was proposed based on color and area characteristics, and a cocoon harvestor was designed based on machine vision. The monocular CMOS camera was firstly used in the algorithm to take image of checker cocooning frame. And the non-measurement distortion correction method was used to correct the image. Secondly, the camera model was calibrated with the internal parameters for the monocular two-dimensional visual measurement system. The image was smoothed via gray and mean shift filter method because the outer floss of the cocoon can cause wrong segmentation of the image in checker cocooning frame image. Then the binary image was obtained by threshold segmentation. Next, the binary image was processed by open operation and area feature extraction method to remove noise region. A part of the smaller noise connected components can be removed by the open operation. The cocoon region can be extracted by the area characteristic when the large area of the connected components can be removed. The center point coordinates of the cocoon region were got by the connected components calibration, and were mapped into the world coordinates through the equation that transformed image coordinates to world coordinates to get the cocoons’ positions in the Cartesian space. Finally, the cocoons were harvested by the cocoon harvestor. According to the experiment, the algorithm had the accuracy rate of 96.88% for the cocoon detection in the checker cocooning frame and less than 6.0mm for the cocoon coordinate, which satisfied the requirement of the location of cocoon harvesting.

Keywords: silkworm cocoon segmentation, image processing, cocoon harvesting, location algorithm, experiment