害虫的准确识别是针对性地施用农药以有效治理虫害的基础,而人工识别的劳动强度大且主观性强.为此,提出了一种利用颜色特征的害虫视觉识别技术. 使用 GrabCut 算法从虫害图像中分割出完整的害虫主体图像并计算其最小外接矩形区域的H/S通道直方图,使用害虫基准图像对其进行直方图反向投影并计算交叉匹配指数. 匹配指数和害虫标签共同组成的特征向量用于训练 C4 .5 分类器. 计算待检害虫图像的交叉匹配指数,输入分类器即可得到识别结果. 实验结果表明:该技术可准确识别菜蝽、菜青虫、猿叶甲、跳甲及蚜虫5 种害虫,准确率达到92%.%The accurate identification of rapeseed pests is the foundation for using the pesticide pertinently .Manual rec-ognition is labour-intensive and strong subjective .The principal part image of the pets was extracted using the GrabCut algorithm and the minimum circumscribed rectangle of the principal part was calculated .Then histogram backprojection in H/S channels was employed between the template images and the rectangle image to obtain the cross matching ratio .The feature vector consist of the ratio and the label of pests was employed to train the C 4 .5 classifier .With the cross matching ratio of the checking image , the C4 .5 classifier may identify the species of the pets .The experiment showed that the pro-posed method may identify five kinds of rapeseed accurately such as erythema , cabbage caterpillar , colaphellus bowringii baly , flea beetle and aphid with the recognition rate of 92%.
展开▼
