Vision Based Estimation Of Motion Parameters And Gradient Of Slope In Craft Landing

The estimation of motion parameters and gradient of slope are important researching contents of a successful landing on the Mars. The estimation of motion parameter is a kind of basis for navigation in Mars landing. The estimation accuracy of motion parameter has a direct impact on the success of the Mars probes landing. It is necessary to improve the estimation accuracy of motion parameters in Mars landing. The slopes are common obstacles in Mars. The accurate estimation of motion parameters is prerequisite to avoid obstacles successfully for the Mars probe. This paper mainly focuses on the estimation algorithm of motion parameters and gradient of slope based on vision in Mars landing.First, the camera imaging model and the algorithm for feature point detection and matching are studied. The author studies the estimate algorithm of the basic matrix F based on feature point matching. The estimation method of the basic matrix F is studied based on Cuckoo search.Second, the author studies the estimate algorithm of motion parameter based on monocular vision. The linear estimation of the rotary matrix R and horizontal movable matrix T are performed by the essential matrix E. The linear estimate values are the initial values for nonlinear optimization of the rotary matrix R and horizontal movable matrix T. the relative attitude angle and displacement are computed by the rotary matrix R and horizontal movable matrix T. the accuracy of algorithm is obtained by simulation in a laboratory and the feasibility of this approach is verified.Finally, the space coordinates of feature points are acquired on the basis of the feature points image coordinates and the feature points match relation. An estimate algorithm of the slope gradient is proposed on the basis of sparse subspace clustering algorithm. The feature points from different slope faces are classified by sparse subspace clustering algorithm and the planes are fitted separately. The normal vector of the two faces is obtained, and further the slope gradient is acquired. At the end, an experiment has been done and the gradient estimation error of this algorithm is lower than 4.82%under laboratory conditions. The feasibility of this algorithm is verified.