Research on image processing algorithm of posture correction of bending larvae

In the process of on-line hazard larva recognition, larvae often appear in bending posture, which makes the extracted feature vectors distorted and affects the recognition results. To solve this problem, this paper proposes an image processing algorithm based on sector-shaped transform to correct the bending posture and improve the larva recognition rate. Firstly, bending larvae are segmented in order to determine the bending region and nonbending region after image thinning, linear fitting and circle fitting. The bending region is corrected to straight line by sector-shaped transform; and the nonbending region and the bending region that has been processed by rotation, translation and sector-shaped transform are combined to form a complete larva. Finally the eight-neighborhood mean method is used to calculate the pixels of gaps in the combined larva. Euclidean distances of seven feature vectors based on moment invariants before and after the posture correction and the standard posture show that the algorithm can greatly improve the similarity of feature vectors between larvae of bending posture and standard posture.