用于柑桔成熟度无损检测的色度频度序列法研究

农产品内部品质无损检测技术是确定水果最合适的采收期和对成熟度不一致的农产品进行准确分级的关键．为了建立利用计算机视觉技术进行柑桔成熟度的无损检测的方法,本研究建立了用于柑桔成熟度检测的计算机视觉系统,研究了柑桔图像颜色的描述方法,通过分析比较,认为在利用水果可见光彩色图像检测水果成熟度时,宜采用HSI颜色模型空间;提出了用与各个色度对应的像素在图像中出现的频度构成的频度序列描述图像的颜色信息的新方法,并利用人工神经网络方法建立了根据柑桔图像的色度频度序列判断柑桔成熟度的映射器,这一映射器检验252只成熟度不同的尾张系柑桔的结果为,对成熟果实和未熟果实的判断正确率分别为 79 1％和63．6％,总的判断正确率为77．8％,这表明尾张系柑桔果实的表皮颜色与成熟度之间具有相关性,可以通过利用计算机视觉技术测定柑桔的表皮颜色信息来判断柑桔的成熟度．

Non-Destructive Maturity Evaluation of Citrus by Hue Frequency Sequence Method

Non-destructive quality evaluation of agriculture products is the key to determine the suitable harvest date and to grade the agriculture product. A computer vision system for maturity inspecting of citrus was set up, and the color describing method of citrus images was studied. It was founded that HSI color space should be adopted to describe the citrus image, while evaluating the citrus maturity by the information in the visible color image. A new method to describe the color information of citrus image was put forward, in which the color information was depicted by hue frequency sequence--the sequence consisted of the number of pixels corresponding to different hues. A triple-layer feed-forward network was established to mapping the maturity of citrus from the hue frequency sequence. After training, the network mapper was used to inspect the maturity of the test sample set, which was composed of 252 Weizhang citrus with different maturity. The identification accuracy on mature citrus reached 79.1%, on non-mature citrus was 63.6%, and the average identification accuracy was 77.8%. It was concluded that the maturity of Weizhang citrus is related to their surface color information, and it is feasible to nondestructively evaluate the maturity of citrus with the computer vision system and hue frequency sequence method.