Abstract: Aiming at the drawbacks that the multiple linear regression model is not ideal to the boundary segmentation for insect image with shadow,a new insect image segmentation algorithm based on multiple linear regression and transition region was proposed in this paper．The method first optimizes the multiple linear regression model with norms,i.e．builds the multiple linear regression model based on RGB three color boards of the image,then acquires the regression model optimized by the cosine norm to segment the insect images．The optimized algorithm can improve the image segmentation effects,but it also retains the image shadow．So the transition region algorithm was introduced to segment the boundary from the shadow,and achieve the boundary optimization．Compared with the algorithm which only uses multiple linear regression,the proposed method improves the accuracy of insect image segmentation and has strong robustness.

Key words: Insect image,Multiple linear regression,Transition region segmentation