基于图像处理与人工神经网络的小麦颗粒外观品质评价方法

综合利用计算机视觉、图像处理、人工神经网络技术，实现小麦品质评价自动化。通过比较不同背景，发现在黑色毛面纸板背景下，使用数码像机获得容易处理的小麦图像。应用分水岭算法自主开发了图像分割处理软件，分割小麦图像并识别提取出完整的小麦颗粒，针对每个小麦颗粒，计算了其12个形态学特征、12个色泽参数等图像特征参数。利用所提取的24个小麦图像特征参数，采用人工神经网络BP算法建立起小麦粒径外观品质评价模型，并应用于小麦的品质识别，取得了良好的试验结果。多次建模运算证明，该方法具有较好的稳定性，对小麦粒径外观品质评价的平均识别准确率可达93%。

Method for identification of external quality of wheat grain based on image processing and artificial neural network

In this paper, computer vision, image processing and BP neural network were combined together to realize automatic identification of external quality of wheat grain. After comparing different background images, it was proved that images taken in a black flock paper background by a digital camera were the most preferable. An image segmenting software based on watershed algorithm was then designed to segment wheat images and identify each grain. For each wheat grain, 12 morphological characteristics and 12 color and luster parameters were calculated. With these 24 features, a wheat quality identification model was developed by applying BP artificial neural network. The model was employed to identify wheat quality. After times of modeling, it was proved that the final model was stable and repeatable. The average identification rate reached 93%.