Image Segmentation Of Touching Rice Based On Machine Vision

As the quality of people’s lives improving and the international grain marketInternationalizing, rice quality detection will directly affect people’s lives and China’s ricecompetitiveness in the international market, it is of great significance. In the past, Most of therice quality detection based on labor, it is lack of accuracy and objectivity. In this paper,touching rice image segmentation based on machine vision, it will provide technical supportfor the rapid and real-time detection of rice quality, and conducive to the establishment ofstandards of rice quality inspection system.On the basis of the analysis of domestic and foreign touching particles segmentation, thispaper proposes a new segmentation of touching. This paper uses tile rice as the researchobject, collect four image from front, left, and right directions of the rice flat under the Fixedlight source height. Adjust the light height images and collect image in accordance with theabove way. Then use the power transform, histogram equalization and local enhancement toenhance the rice image. Then use morphological to denoise the binarization image. In thesegmentation step, firstly, use LoG, Canny, Prewitt, and zero-crossing operator to extract theedge of the image on the same height. And sum the four edge to get one edge. Get a completeand continuous edge by using morphological expansion and refinement. Lastly, contrast theedge and the original image to implement the segmentation. The experimental results showthat: under suitable light source height conditions，This method can effectively segment thetouching rice with different shape and different touching condition. The effect is quitesignificant. Compare with the traditional method, this method the advantage: the algorithm issimple and easily implemented in hardware.