Object Detection Optimization Based on Prior Feature Clustering
In addressing the issue of salient object detection without any prior information,this study proposes an object detection optimization method using feature clustering and a compactness prior scheme.The optimized approach consists of four steps:Firstly,a superpixel preprocessing is employed to segment the image into superpixels,suppressing noise and reducing computational complexity;Secondly,an improved shrimp swarm clustering algorithm is applied to classify color features;Additionally,two-dimensional entropy is utilized to measure the compactness of each cluster and to construct a background model;Finally,the contrast between the background region and other regions is used as a salient feature,enhanced through the design of Gaussian filters to amplify its saliency.To better evaluate the accuracy of salient object detection,this paper conducts experiments and analysis using multidimensional evaluation metrics,and the results demonstrate that the proposed algorithm exhibits good real-time performance and robustness.
万方数据
维普
