关键词: 甘蔗, 复杂环境, 蔗梢图像分割, 惯性权重, 非对称加速因子

Abstract: In order to realize the effective identification and cutting of sugarcane tip by sugarcane combine harvester, a sugarcane tip image segmentation method based on improved particle swarm optimization algorithm combined with OTSU threshold segmentation algorithm was proposed. Taking the H component of sugarcane image as a segmentation sample, the median filter was used to remove small range of isolated noise point in the image, and the PSO algorithm with improved asymmetric acceleration factor and update nonlinear decreasing inertia weight was used to  iteratively search for the optimal threshold of the image, and the threshold was used as segmentation threshold of the improved algorithm, and sugarcane tip segmentation image was obtained. According to different weather conditions, 120 sugarcane sample images were collected, and the algorithm was tested. The experiment showed that the segmentation and recognition rate of sugarcane image under sunny and cloudy weather was 88.33% and 91.67%, respectively, and the average recognition rate was 90.0%. The average processing time of single sugarcane image was 0368 7s. The segmentation speed was 30% and 15% higher than the traditional OTSU algorithm and the standard PSO+OTSU algorithm, respectively.

Key words: sugarcane, complex environment, cane tip image segmentation, inertia weight, asymmetric acceleration factor