稻飞虱图像形状和颜色特征提取的研究

Extracting Shape and Color Features of Rice Hopper Images

为实现灯下三种稻飞虱的自动识别和计数,对稻飞虱图像特征提取进行研究。通过对稻田飞虱人工识别特征的分析,确定采集稻飞虱的背部图像。在提取背部特征的研究中,首先对6张稻飞虱单样本图像由傅立叶系数提取形状特征不变量,采用色度百分率直方图法提取背部色度最大值。然后计算大量稻飞虱样本形状和颜色特征值的标准方差以确定各特征值的稳定性。结果表明,白背飞虱图像的各特征值最为稳定。

In order to realize the automatic recognition and count of rice hopper, this paper studied the extration of shape and color features from rice hopper images. Through analyzing the artificial reeognitive features, the back images of rice hopper were aequisited. During the study on extracting back features, the shape feature invariant of 6 single sample images was firstly extracted by Fourier coeffieient, and color feature was extracted by ehroma percentage histogram; then the stability of shape and color feature was confirmmed by computing the standard deviation of the feature values of a lot of sample images. The results showed that the features of Sogatella furcifera images were the stablest.