Landing Cooperative Target Robust Detection via Low Rank and Sparse Matrix Decomposition

摘要

The premise and foundation of autonomous landing relative navigation based on vision for unmanned aerial vehicle (UAV) are accurate detection of landing cooperative target. In order to overcome the problem of cooperative target detection susceptible to illumination change and the interference, a new landing cooperative target detection algorithm based on low rank matrix recovery theory is proposed. In our model, an image is represented as a low rank matrix plus sparse noises, where the landing cooperative target can be explained by the sparse noises, and the background is indicated by the low rank matrix. By giving an image, we extract visual low-level features and higher-level to construct the image feature matrix. Then this model is decomposed to be a low rank matrix and a sparse matrix by robust principal component analysis (RPCA). Finally, the image patch with the highest saliency in sparse matrix is extracted and considered to be the landing cooperative target. Extensive experiments show that the proposed method has a high detection accuracy and good robustness.