基于颜色特征的马铃薯绿皮检测

Greened Surface Detection of Potatoes Based on Color Character

提出了一种基于颜色特征的马铃薯绿皮检测方法,该方法以颜色特征分析为基础,结合BP神经网络,实现对马铃薯绿皮缺陷的检测。首先建立马铃薯图片采集系统,采集丰富的绿皮马铃薯和优质马铃薯图像并对图像进行彩色图像增强。对绿皮缺陷和优质区域的颜色特征进行深入分析,提取出相对独立的特征参数R、G、B、H、R/G、B/(R+G)。构建了BP神经网络,以特征参数作为输入对马铃薯图像像素进行判别,实现绿皮区域的准确分割。实验结果表明该方法能够正确识别绿皮区域,准确率达96.88%。

A method for greened surface detection of potatoes based on color character was put forward.This method is based on color character analysis and combines with neutral network to realize the detection of greened surface of potatoes.At first,image collection system was set up and certain number images of greened potatoes and normal potatoes were gathered and enhanced using color image enhancement method.Then deeply research the difference of color character between greened surface and normal surface and abstract R、G、B、H、R/G、B/（R＋G） was then carried out.Finally was the building of a BP neutral network to identify the greened pixel of the potato image and realize the accurate segmentation of the green area.The result of experiment showed that this method can accurately recognize the green surface and the accurate rate can reach 96.88 percent.