Methods And System Implementation On Three-Dimensional Attitude Measurement Based On Monocular Vision

With the maturation of machine vision and the rapid development of visual sensors towards to high-speed, high accuracy and high sensitivity, monocular vision measurement technology are being used more extensively. It has been already the new star at the fields of aerospace, industrial automation, artificial intelligence and so on. The purpose of this paper is to measure the attitude of the three-dimensional moving target based on monocular vision. The measurement systems based on monocular vision includes the study of the Image Processing problems such as image preprocessing, camera calibration and so on. Furthermore, this paper compares and analyzes different attitude solution algorithms based on monocular vision and then built the hardware and software platform.First of all, the paper defined the camera coordinate system, the target coordinates and image plane coordinate system. The relationship between the coordinate systems were analyzed, then the ideal model of attitude measurement system was established based on monocular vision and the principles of attitude measurement based on monocular vision was studied.Secondly, describe the principles of feature points , according to these principles, design the non-coplanar feature points target to improve system accuracy then the camera calibration algorithms was studied, in considering the specificities of camera and target, the paper chooses the two-step calibration method to do camera calibration. Furthermore, image pre-processing process and principles, the principles and algorithms of feature points positioning were introduced. Image preprocessing, includes two parts of image filtering and segmentation. Depending on the camera to obtain images easy to produce isolated point noise and blurred edges, the paper selected median filter and verify the actual effect of image median filtering. Since then, the technology of threshold was used to devise the image feature points and background. The algorithm of feature points positioning was fast, simple and also owned high accuracy.Then, the linear attitude solution algorithm and iterative algorithm based on a total of linear equations was studied. Simulation experiment was done to clarify which situation the two algorithms can be used based on clarifying the principle of two algorithms. In summary, iterative algorithm is fit for our systems, and then the overall implementation scheme about attitude measurement system based on monocular vision was determined.Finally, the software and hardware platform of attitude measurement system base on monocular vision was set up. With the existing camera control software and image acquisition software, this paper development the image storage and attitude solution software. After that, the experiment of attitude solution was done and then the experimental results and the reasons of producing errors were given.