Rice planthopper segmentation based on multi-feature fusion

Diagnosis of rice planthopper pests based on imaging technology is an efficient means to develop intelligent agriculture. Effective contour automation extraction is an important pretreatment technology at the early stage for identifying and classifying rice planthoppers. For the curtain as the background contained texture structures and resulted in a heterogeneous texture in the sensed image, which made the rice planthopper segmentation extremely difficult, this paper presented a new method, based on digital image processing, to segment rice planthoppers on curtain in paddy fields. First, based on a nonlinear dispersive phase operation applied to the image, the output phase of the transform was used for edge detection. Second, a multi-feature fusion mean shift method based on color features, and edge features which were fused to enhance the target information to reduce the localization error of object segmentation was proposed. Experimental results had shown the proposed method achieved success in rice planthopper segmentation, at the same time, feature fusion strategy made the algorithm more robust, especially with the changes in the illumination condition.