| The UAV can take the place of people in danger and extreme environments may minimize the loss of people and reduce personnel costs, so that it has a rapid development in recent years. In military reconnaissance, remote sensing, anti-disaster insurance, pipeline patrol, aerial video etc. has been widely used. While the UAV visual navigation as the basis of UAV automatic path planning, real-time target tracking and multi machine collaboration work, has become the focus of current research.This paper design and implements an UAV visual stable obstacle avoidance system, through the UAV airborne information acquisition system and PC ground station feedback control UAV system components, in which the airborne information collection system is divided into the integrated navigation information system and image transmission and storage system. By dividing several subsystems, the whole system is independent of each part, equipped with integrated flexible and transplants each part conveniently.For airborne image acquisition system, this paper completed by translational compensation of real-time electronic image stabilization system has been improved for the laboratory, designs a kind of airborne electronic image stabilization system based on affine transformation, which can be realized on the translation and rotation superimposed video dithering process. Based on the FAST feature point matching by affine transformation model parameters are estimated. Through the moving average filter separation jitter and scanning motion, finally by combining the method of space transform summation and nearest neighbor interpolation to complete the motion compensation of the image. The system uses ADSP-BF609 symmetrical dual-core 500 MHz processor core, optimized design of parallel algorithms and systems. For PAL standard video processing system, the system has real-time requirements, and reliable.For the PC back control of UAV ground station, this paper design a method by translation to measuring the distance of monocular fixed focus image, based experiments show that the method is simple, and the accuracy to meet system requirements. According the UAV distance to calculated the projected area, then combined with the detection method of AABB bounding box collision to accurately determine whether the intersection of UAV projection with obstacles. For rotorcraft UAV capable of vertical flight characteristics, through the improved threat cone model for rotorcraft UAV features, so that the UAV to recent collision point as the center to avoid obstacles, ensure the effective avoidance of UAV.Finally,testing calibration system MEMS inertial devices, navigation information collection, airborne image capture output. And electronic image stabilization system and monocular obstacle avoidance control system back to experimental analysis, in order to verify integrity and real-time systems and the effectiveness of the algorithm. |
