基于三摄像系统的苹果缺陷快速判别

Fast identification of apple defect based on three imaging system

水果缺陷识别一直是计算机视觉水果实时分级中的研究热点，文中提出一种基于三摄像系统的逻辑判别缺陷的新方法．首先利用三摄像系统获取苹果在三个连续不同位置的9幅图像，然后对图像进行去背景、滤噪等预处理操作，再对图像进行缺陷的分割并加以标记，接着根据分割标记后的可疑区个数，对苹果缺陷的有无进行逻辑判别，即当这9幅图像中只要有一幅图像被分割出两个以上的可疑区，则判断该苹果有缺陷．试验结果表明，缺陷识别精度达到89．4％，另外由于该方法避免了缺陷和果梗或果萼直接识别，处理速度和正确率都得到了很大的提高，能够满足在线5—10个／秒的检测速度要求．

Identification of fruit defect is an attractive subject in on-line detection of fruit quality. A new logical identification method of fruit defect based on three imaging grabbing system is put forward. First, nine images of an apple is taken at three continuous positions by the system. Second, the apple is segmented from the black background by thresholding and smoothing. Third, defect＇s segmentation and counting are performed on the apple＇s images. Last, logic distinguishment was performed, then an apple has defects if one of the apple＇s nine images has two doubtful regions. The result demonstrated that the new method is very useful and effective, and the accuracy of defect identification is up to 89. 4%. The system can meet the requirement of 5 ～ 10 apples every second in on-line detection.