基于HOG特征的IKSVM稻瘟病孢子检测

为解决稻瘟病孢子的人工检测过程中主观性强、自动化程度低、效率低等问题，提出一种基于梯度方向直方图特征（HOG特征）的加性交叉核支持向量机（IKSVM）的稻瘟病孢子检测方法。该方法首先利用图像采集系统采集稻瘟病孢子图像，利用Gamma校正法调节图像的对比度，抑制噪声干扰；然后，提取孢子图像的HOG特征作为输入向量，输入到支持向量机中，构建加性交叉核支持向量机分类器；最后，通过训练得到稻瘟病孢子分类器。为测试所提出的HOG/IKSVM方法的综合性能，分别选用HOG/线性SVM方法与HOG/径向基核SVM（HOG/RBF-SVM）方法做对比试验。试验结果表明，HOG/IKSVM的检测率为98.2%，高于HOG/线性SVM方法的79%；在平均检测时间上，HOG/IKSVM方法的平均检测耗时仅为HOG/RBF-SVM方法的1.1%。说明该方法可以进行稻瘟病孢子室内检测识别。

Spores Detection of Rice Blast by IKSVM Based on HOG Features

In order to solve the disadvantages such as strong subjectivity, low automation and low efficiency of spores detection in rice blast, an additive intersection kernel support vector machine (IKSVM) based on histogram of oriented gradient feature (HOG feature) was proposed to detect rice blast spores. Firstly, the image acquisition system was used to collect spores images of rice blast disease, and Gamma correction was used to adjust the contrast of the images to suppress noise interference. Secondly, the HOG feature of the spores image was extracted as input vectors and input into the support vector machine to construct the intersection kernel support vector machine classifier. Finally, the rice blast spores classifier was obtained by training. In order to test the comprehensive performance of proposed HOG/IKSVM, the HOG/linear SVM method and the HOG/radial basis function kernel SVM (RBF-SVM) method were used for the comparison test. The test results showed that the detection rate of HOG/IKSVM was 98.2%, which was higher than the 79% of the HOG/linear SVM method. On average detection time, the average detection time of HOG/IKSVM was only 1.1% of the HOG/RBF-SVM method. This method can be used as a rapid and accurate identification method for indoor detection of rice blast.