Target-oriented deformable fast depth estimation based on stereo vision for space object detection

作     者：Xu, Chengcheng Zhao, Haiyan Gao, Bingzhao Liu, Hangyu Xie, Hongbin 

作者机构：Jilin Univ Coll Commun Engn Changchun 130025 Peoples R China Tongji Univ Sch Automot Studies Shanghai 201804 Peoples R China 

出 版 物：《MEASUREMENT》 (Meas J Int Meas Confed)

年 卷 期：2025年第245卷

核心收录：

学科分类：08[工学] 080401[工学-精密仪器及机械] 0804[工学-仪器科学与技术] 081102[工学-检测技术与自动化装置] 0811[工学-控制科学与工程] 

基　　金：National Nature Science Foundation China Jilin Province Science and Technol-ogy Plan Program [20230201123GX] 

主　　题：Intelligent vehicles Space object detection Object matching Depth estimation 

摘      要：To address problems of the poor matching accuracy and speed in space object detection, this paper proposes a deformable fast object matching algorithm. It is a target-oriented depth estimation approach that calculates the disparity value by matching the object on the left and right images. The logical encoding masking layer is designed to achieve the deformable operation, which can fully fuse the semantic or contour feature information of the object. This effectively reduces the computational cost and improves the accuracy. The feature coding method is optimized and upgraded by integrating relative positions and global pixels in the image, solving the problem of mismatching in complex regions without obvious features. Based on the characteristics of stereo vision and the concept of regional matching, an optimal matching search range is proposed. Results show that the average time is less than 9ms and accuracy reaches the state-of-the-art.