Aiming at the problem that the ineffective area and flat area in the continuous local gear image will affect the registration time and registration accuracy of the feature-based algorithm

a mosaic algorithm for gear image based on region segmentation and A-KAZE is proposed.Firstly

the algorithm uses the area blocking method and relevance measure to identity the similar regions between the reference image and the image quickly and the feature region.Then

in the similar regions

the A-KAZE algorithm for image feature extraction and description is used.Next

the KNN algorithm based on the Hamming distance roughly matches the feature points

and RANSAC algorithm calculate the best geometric transformation model.Finally

the weighed average method is used for image fusion.The experimental results show that the algorithm reduces time in the invalid area and flat area

and has high precision.It has better performance than the traditional algorithm

and is more suitable for continuous local gear image stitching.