一种改进SM谱聚类算法的棉田棉花精确分割

An Improved SM Spectral Clustering Algorithm for Accurate Cotton Segmentation in Cotton Fields

根据棉田棉花生长特点,利用最大熵法进行背景增强,在Lab颜色空间中进行图像预处理;利用粒子群优化算法得到局部最优,结合SM谱聚类算法,自发式地快速确定最佳聚类数。对自然环境下棉田棉花图像的进行分割实验,识别准确率为93.1%,相对误差率为2.22%,分割时间为2.067s,证明所提出的算法对于复杂环境下的棉花图像分割的精确性较高,并能够满足采棉机器人在实时行走过程中对图像分割的准确率和实时快速的要求,为采棉机器人的路径规划提供了快速、有效的新方法。

According to the characteristics of cotton growth in cotton field,the maximum entropy method is used to enhance the background,and image preprocessing is carried out in the lab color space;the particle swarm optimization algorithm is used to obtain the local optimum,and the SM spec-tral clustering algorithm is used to quickly deter-mine the optimal cluster number.Through the seg-mentation experiment of cotton image in cotton field in natural environment,the results show that the accuracy of cotton recognition in cotton field is 93.1%,the relative error rate is 2.22%,and the seg-mentation time is 2.067s,which proves that the proposed algorithm is fast and accurate for cotton image segmentation in complex environment,and can meet the requirements of image segmentation accuracy and real-time speed of cotton picking ro-bot in the process of real-time walking.It provides a fast and new solutions for the path planning of cottonpicking robot.