介绍了苹果颜色自动分级系统的硬件组成,确定了苹果颜色特征的提取方法,利用遗传算法实现了多层前向神经网络识别器的学习设计,实现了苹果颜色的实时分级,并通过实验验证了方法的有效性.试验结果表明,颜色分级识别准确率在90%以上,分级一个苹果所用的时间为150ms.

Color is one of the most significant inspection criteria related to apple external quality. In this paper, a computer vision experimental system for apple color estimation was first discribed. The system included a single-lane conveyer, an apple illumination chamber, and the hardware fot apple image acuquisition and procession. A method of using HSI color system and neural network techniques for apple color inspection was developed. A GA-based training algorithm was introduced to find optimal structure or the number of hidden layer nodes and connection weghts of artificial neural network. The results of experiment show that the approach is effective fot real-time color grading and is accurate. The vision system achieved over 90% accuracy in color classification for apples by representing features with hue histograms and applying artificial neural network. The executing time of microcomputer for grading of one apple is 150ms.