棉花叶部害螨病斑形态特征提取

Shape feature extraction of leaf lesion for cotton mite disease

针对当前农作物病害诊断存在的效率较低、难以保证精确度等问题,提出运用计算机图像处理技术进行棉花害螨病斑特征提取的方法.该方法以棉花害螨病叶为研究对象,利用中值滤波法对噪声的干扰进行有效的去除;结合运用超绿特征2G-R-B分割算法和面积阈值法将害螨病斑区域从病叶图像中有效分离出来;最后依据分割好的病斑样本图像,运用二值图像区域标记法准确提取出病斑的8个形状特征值.对提取的数据进行分析,得出病斑的圆形度、伸长度紧凑度和内切圆半径等4个相对值特征能有效地体现病斑的形状特征,可以作为识别病害症状的依据.试验结果表明,该方法准确有效.

In view of the inefficiency and difficulty to guarantee the accuracy of the current crop disease diagnosis ,the feature extraction method was put forwarded using computer image processing technology of cotton mite disease spot. The method used the infected leave of cotton red mites as the research object and used median filtering method to effectively remove the interference of noise;then the mite lesion area was effectively separated from the diseased leaves image by combining the algorithm of excess green feature 2G- R-B and the area threshold method. At last, the eight shapes of values which related to the lesion area were extracted through the use of the binary image area notation according to the split lesion samples. Based on the analysis of the data obtained from spot,4 relative value features such as roundness ,elongation of com- pactness and inscribed circle radius can effectively reflect the shape features of lesion which can be used as one of the basis for identification of disease symptoms. The experiment results showed that the area of cotton red mite lesion could be effectively divided and the shape features of the mite lesion were accurately extracted, which could lay the foundation for the disease identification and diagnosis.