关键词: 机器视觉, 图像处理, 金银花识别, Canny算法, 边缘检测

Abstract: The automatic picking of honeysuckle can be realized using the methods of machine vision and image processing, and the picking efficiency can be improved. Firstly, the image of the honeysuckle was collected by cameras, and the collected honeysuckle image was processed by a median filter to eliminate the noise in the image effectively. Then, RGB and HSV color segmentation were carried out on the honeysuckle image to find out the component B that distinguishes the honeysuckle from the background most obviously. Then the threshold segmentation of component B was processed, the threshold was set, the honeysuckle was extracted from the background, and the morphological operation was used to make the image fuller. Finally, the Canny algorithm was used to study the edge detection of honeysuckle images, and the Canny algorithm was improved to achieve a better edge detection effect. The results showed that the recognition rate of honeysuckle through threshold segmentation was 79.17%, the recognition rate of the traditional Canny algorithm was 66.67%, and the recognition rate of the improved Canny algorithm was 93.75%, which can meet the realtime operation requirements of subsequent honeysuckle picking robot.

Key words: machine vision, image processing, honeysuckle distinguish, Canny algorithm, edge detection