复杂环境下蔗梢图像分割方法研究

Study on image segmentation method of sugarcane tip in complex environment

为实现甘蔗联合收割机对田间甘蔗蔗梢的有效识别与切割,提出一种基于改进粒子群优化算法结合OTSU阈值分割算法的蔗梢图像分割方法。以甘蔗图像的H分量作为分割样本,选用中值滤波去除图像中存在的小范围孤立噪声点,利用改进非对称加速因子与非线性递减更新惯性权重的PSO算法,通过迭代搜寻图像最优阈值,并将此阈值作为改进算法的分割阈值,得到蔗梢分割图像。根据不同的天气条件采集120幅甘蔗样本图像,对本文算法进行测试。试验表明,所设计的算法对晴天、阴天两种天气下的甘蔗图像分割识别率分别达到88.33%、91.67%,平均识别率为90.0%,处理单张甘蔗图像平均所用时间为0.368 7 s,分割速度较传统OTSU算法与标准PSO+OTSU算法分别提高30%、15%以上。

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 7 s.The segmentation speed was 30% and 15% higher than the traditional OTSU algorithm and the standard PSO+OTSU algorithm,respectively.