On Mars Surface Image Acquisition And Processing Method For Optical Navigation

Mars exploration is a hot issue of deep space exploration in various countries.Since Mars is far away from the earth,there should be highly real-time requirements of the probe navigation,which is a challenge to the autonomous navigation technology.Using surface image information as a data source to carry out autonomous navigation has shown the advantages of convenience and rapidity.In this thesis,we take the Mars exploration as the background,and perform autonomous navigation oriented image acquisition,optimization,and matching in optical navigation.We aim to improve the quality of the acquired images,obtain good matching datas,and provide technical supprot for subsequent positioning work.(1)The optimization of the acquired image is carried out.To address the Gaussian noise interference in the Martian environment and probe acquisition equipment,this thesis proposes a Gaussian difference denoising scheme based on feature detail preservation.By using the two-step processing of noise removal and detail enhancement,this scheme improves the problem of detail loss in traditional Gaussian filtering while ensuring the rapidity of the algorithm.To address the poor brightness and low contrast of the collected images due to changes in the Martian environment,this thesis proposes a histogram equalization scheme that integrates the characteristic information of the Martian surface.The scheme performs edge information enhancement processing based on the "sobel+gamma",and limited contrast equalization processing based on "CLAHE" on the collected images.This scheme avoids the brightness overexposure in traditional equalization and highlights the details of surface features.This provides data foundation for the subsequent feature extraction and matching work.(2)The image feature extraction and matching are performed.Firstly,considering that the number of detection features of the ORB algorithm is small and easy to cluster when processing low-textured Martian surface images,this thesis uses nonlinear scale space to replace ORB scale space,and designs the OFAST feature detection scheme with an adaptive threshold.Then this thesis adopts the BRISK-FLANN method to perform feature matching and designs a MAGSAC mismatch elimination scheme based on prescreened point sets.The scheme achieves the purpose of mismatch elimination through initial point set screening and MAGSAC optimization.Finally,this thesis designs a local feature point matching strategy based on the region area,which further improves the matching speed.(3)This thesis builds a detector image acquisition system based on FPGA+DDR3,and designs the clock module,camera configuration module,data integration module,and DDR control module.The system is used to collect the Martian sand table.We then perform the optimization and feature matching on the acquired images.The effectiveness of the acquisition system is verified,demonstrating the effectiveness of the mars image optimization scheme and feature matching scheme.