The Study Of Target Positioning Technology Based On UAV Image Sequence

Being a economical, safe, rapid and efficient platform of acquiring ground information, Unmanned Aerial Vehicle(UAV) has been noticing widely and researching deeply in country investigation, resource exploration, environment inspection, communication relay and battlefield ground information acquirement. Because of the airflow's serious influence and the irregular structure of the cameras fixed in UAV, the images obtained by UAV have incline, rotation, zooming and distortion, which cause low usage factor and difficulties in image matching, panorama generation and moving target positioning. This thesis makes deep and particular research and practice to solve these problems, and the main works and productions are listed as follows:1. On the basis of expounding the imaging precess and influence factors of non-metric cameras, various calibration methods are introduced and analysed in this thesis, and a method with plane template which is easy to operate and has high precision is chosen to work out the internal parameters and lens-distortion parameters. The parameters are necessary in target positioning and distortion rectification.2. Through a mass of experiments, the relationships between the camera internal parameters and the image's focus value are found, and a method of zoom camera calibration is proposed. This method only need to calibrate several focuses to work out unknown coefficients, and the internal parameters and distortion parameters corresponding with other focus values can be interpolated to obtain. This method is wieldy and practical.3. According to the characteristics of UAV image sequence, we extract image features with SURF operator, and optimize the matching arithmetic to satisfy the accuracy and speed need. Prompt matching and mosaicing of image sequence in the improved arithmetic described in the thesis are implemented. The panorama generated in the method of the thesis could generally and directly reflect the ground information of the goal areas and is convenient for user to browse every region and key target.4. As the images obtained by UAV can't satisfy a series of conditions required in photogrammetric positioning technology, and considering the flexibility and fault tolerance of the computer vision positioning technology, a method for moving target positioning based on monocular vision is proposed. With the internal parameters supplied by calibration and the external parameters carried out from POS data, positioning for moving target on the ground is implemented. Compared with the traditional photogrammetric method, the method of this thesis is more flexible and rapid, and promotes the appliance efficiency of UAV.