A feature‐assisted matching algorithm for lunar rover

Abstract

PurposeTo present a stereo matching algorithm which satisfies the need of visual navigation on outdoor natural terrain for lunar rover or other mobile robots.Design/methodology/approachA feature‐assisted matching algorithm is presented to generate dense and accurate disparity map of natural terrain. Multi‐feature matching strategy produces reliable matching results for edge points. Disparity monotony constraint is derived and other geometrical constraints are introduced. With these constraints the edge‐matching results are used to limit the search region in area‐matching. As a result the algorithm produces dense disparity maps with fairly high accuracy and demonstrates advantages over straightforward area‐matching algorithm in improving matching accuracy.FindingsWith the help of several constraints, the feature‐assisted matching algorithm performs well in the matching of stereo image pairs of natural terrain.Research limitations/implicationsThe algorithm focus on improving the accuracy of stereo image pairs matching of natural terrain and computation complexity is not an important designing factor. Only with the assistance of special hardware or other technique can the algorithm be used for real‐time navigation.Practical implicationsThe algorithm is able to produce dense disparity map of natural terrain with rather high accuracy and can be used for the navigation of lunar rover or other outdoor mobile robots.Originality/valueThe paper provides a new approach to produce accurate and dense disparity map of natural terrain.