论文部分内容阅读
为了提高玉米品种识别的准确率,提出了一种基于深度和颜色的灰度直方图结合BP神经网络的玉米品种分类方法。使用深度传感器获取玉米子粒的深度图像,并将获得的RGB彩色图像转化为HSV图像进行分析,发现不同品种的H分量有明显差异,从而确定不同颜色范围对应的灰度值,用归一化和灰度化后的图片生成灰度直方图,发现不同品种的灰度特征值差异比较大,取其中重要的4个灰度特征值作为BP神经网络的输入,经过训练识别出不同的品种。试验结果表明,此方法识别出的玉米品种与人眼观察的结果基本一致。
In order to improve the accuracy of maize variety identification, a maize variety classification method based on depth and color grayscale histogram combined with BP neural network was proposed. Using depth sensor to obtain the depth image of maize grain, and converting the obtained RGB color image into HSV image for analysis, it is found that the H components of different varieties have significant differences to determine the corresponding grayscale values in different color ranges. Using the normalized sum Grayscale images generated histogram and found that different varieties of gray-scale eigenvalues are relatively large, take the four important gray-scale eigenvalues as BP neural network input, after training to identify different breeds. The test results show that the corn varieties identified by this method are consistent with those observed by human eyes.