Mono-vision Based Aerocraft Pose Measurement Technology

Vision photogrammetry (VP) is an advanced and useful primary system based on computer vision theory, which applies low noise & distortion imaging sensors, high speed real-time imaging acquisition system, dedicated imaging hardware processing system and high performance computers to realize photogrammetry. It booms for the maturation and integrity of the electronics, photodetection, imaging processing and computer techniques, and it is widely used in range testing, satellite control ect. Currently the emphasis of VP is placed on the estimation of objects'size, position and pose. VP can be divided into three classes by the imaging sensors'number, namely, mono-vision photogrammetry, stereovision photogrammetry, and multi-vision photogrammetry. Mono-vision photogrammetry uses only one camera or vidicon to implement photogrammetry. For this reason, the framework of the method is simple, and the calibration is easy. At the same time, it avoids the problem of small visual field and the stereo matching which exists in the stereovision and multi-vision.Based on the above, the paper investigates the missile's pose estimation by mono-vision, which is derived from the project,"Mono-vision based aerocraft pose estimation". The main work and contribution of the paper are as follow:1. Through wide reading, the previous studies is comprehensively reviewed, and the existing mono-vision photogrammetry methods is also summarized. A mono-vision photogrammetry scheme for aerocraft is given in the paper. In the scheme, we use the geometry feature points as the control points, and needn't mark the object. It can acquire the correspondence of the object control point sets automatically by feature points extraction and point pattern matching. After these, we obtain the estimation of the object's pose by Kalman Filter. The scheme is flexible, universality and expansile.2. An ant colony optimization (ACO) based approach for point pattern matching (PPM) is proposed in the paper. In the method, the point sets matching problem is formulated as a mixed variable (binary and continuous) optimization problem. The ACO is used to search for the optimal transformation parameters. There are two contributions made here. Firstly, We manage to modify the original ACO method by combining it with the least-squares method. Thus, it can handle with the continuous spatial mapping parameters searching. Secondly, we introduce a threshold to correspondence finding, which rejects outliers and enhances veracity while using"Nearest Neighbors Search". Experiment results demonstrate the validity and robustness of the algorithm.3. PNP problem is studied in the paper. Pose estimation of the aerocraft is a typical PNP problem while the camera is calibrated. When N≥6, PNP can turn into a linear problem. When N<3, there is no solution. While 3≤N≤5, PNP have several solutions. The existing algorithms are all non-linear methods, and they are quite sensitive to position error of the image point, which can't be avoided in reality. Therefore, P3P and P4P problem is emphasized on in the paper.4. An Kalman Filtering method is presented to obtain the estimation of the object's pose through image sequence. Simulations results show that the precision of the pose estimation is greatly improved, and the method is effective and applicable.