针对最大类间方差法在图像分割时存在造成噪声干扰和过分割的缺点,提出一种基于改进和声搜索算法的玉米叶片病害图像分割算法。算法将玉米叶片病害图像编码处理,选取图像的类间方差作为改进和声搜索算法的适应度值,通过改进和声搜索算法寻找最优的分割阈值,利用该最优阈值使用经典最大类间方差法对玉米叶片病害图像进行分割。选取强光、中光、弱光条件下三幅玉米叶片病害图像进行分割实验,结果表明采用基于改进和声搜索算法的玉米叶片病害图像分割算法较最大类间方差法和基于混合蛙跳算法的图像阈值分割算法均具有较好的图像阈值寻优能力,可有效提高玉米叶片病害图像中病斑分割的效果。
To solve the problem of noise interference and over-segmentation the OTSU method has in image segmentation,we proposed maize leaf disease image segmentation algorithm which is based on improved harmony search algorithm. By coding the maize leaf disease images,the algorithm selects between-class variance as the fitness value of the improved harmony search algorithm,searches optimal segmentation threshold through improved harmony search algorithm,and uses this optimal threshold and OTSU to segment maize leaf disease images.By choosing three maize leaf disease images under three conditions of strong light,medium light and weak light for segmentation experiments,the results indicated that the maize leaf disease image segmentation algorithm based on improved harmony search algorithm possessed much better image threshold optimisation performance compared with both the OTSU and the image threshold segmentation algorithm based on shuffled frog leaping algorithm,it could effectively improve segmentation effect of maize leaf disease images.