[Objective] Hie aim was to obtain better apples' color image segmentation method under natural environment. [Methods]The situation of blocking (mainly in leaves) and overlapped apples were selected as test picture, the apples' color image was preprocessed by Euclidean distance based on sample, during the pretreatment, the similar criterion of boundary box was used and it improved the processing speed; then black and white' s encoding ( apple and non-apple) resulted in a binary image. Finally the binary image was processed by mathematical morphology. [ Result ] The experiment effectively removed image' s noise which resulted from complex background, targets were obscured and overlapped, demising effect was good. [Conclusion] Experimental results show that the algorithm is fairly efficient for mature apples' segmentation, but not for immature apples.%[目的]获得自然环境下较好分割苹果图像的方法.[方法]选取苹果被遮挡(主要是叶子)、相互重叠的情况作为试验图片,采用欧几里得距离对彩色图像进行预处理,在预处理中,利用边界盒的相似性准则,提高了处理速度;对预处理的结果进行黑白(苹果和非苹果)编码,产生一副二值分割图像;最后采用数学形态学对二值化图像进行处理.[结果]有效的去除了复杂背景、目标物被遮挡、相互重叠等存在的图像噪声,去噪效果良好.[结论]算法对成熟苹果具有很好的分割效果,但不适合未成熟苹果的分割.
展开▼