Research on Image Segmentation Algorithm Based on Combination of SLIC and APMMD
A superpixel is a small region consisting of a series of pixel points with similar features and adjacent positions.The use of superpixel segmentation can both reduce the complexity of image segmentation and better preserve local information and edge information.For the requirements of tube wall stain recognition and parts box part classification,SLIC is used in combination with the nearest neighbor propagation based maximum minimum distance algorithm(APMMD)to achieve better image segmentation results.The method optimizes memory by data type conversion during color space conversion according to the steps of traditional SLIC algorithm.Before uniform assignment of initial seed points,the APMMD algorithm is added to solve the problem of unreasonable initial clustering centers leading to locally optimal clustering results,which can prevent seed points from falling on contour boundaries with large gradients within a certain range.The proposed algorithm is verified by the boundary recall rate and under-segmentation error rate,and it is found that its memory is reduced by 1.5 M during color space conversion,and the accuracy rate is improved by 8.1 percentage points during clustering.
国家科技期刊平台
NSTL
万方数据
