基于CLAHE和开闭运算的绿色苹果图像分割

依据绿色苹果图像的自身特点,设计了一种分区域提取而后合并的图像分割方法。首先对图像进行限制对比度自适应直方图均衡化(contrast limited adaptive histogram equalization,CLAHE),增大果实和背景的颜色差,而后获取R-B色差图像,得到以光常区域为主的区域；然后用CLAHE处理后的图像进行开闭运算,提取局部极大值,得到以高亮区域为主的区域；最后将两区域合并获得完整的果实目标区域。为了验证该方法的有效性,运用Hough算法检测圆,并用相对偏差、圆心相对误差及半径相对误差三个评价指标来定量评价。试验结果表明,提取区域与果实目标区域相比,顺光下,相对偏差、圆心和半径相对误差平均值分别为3.59%,4.76%和2.60%；逆光下,三个指标的平均值分别为10.77%、16.77%、11.49%。无论是顺光还是逆光都有很好的识别效果,能满足机器人采摘果实的精确定位的要求。

Segmentation Method of Green Apple Image based on CLAHE and Open and Close Operation

According to the characteristics of green apple images, this paper designed a segmentation method which is merged after extraction. First of all, the contrast limited adaptive histogram equalization (CLAHE) algorithm was used to increase the color difference between the fruit and the background , and then obtained R - B image. The normal light region of fruit was extracted. Reoccupy open and close operation was processed after CLAHE, to extract the local maximum values, and highlighted area was extracted. Finally full fruit target area was obtained by merging the two regions . In order to verify the effectiveness of the method, hough algorithm is used to detect circle, and three evaluation indexes is applied to quantitative assessment of the method , which are the relative error, relative error of the circle and relative error of the radius. This method has good recognition effect under both the frontlighting and backlight , and it can satisfy the requirement of precise localization of robot.