Font Size: a A A

Key Technology Research On Vision Perception In Unmanned Aerial Vehicles Cooperative Formation Flight

Posted on:2009-08-30Degree:DoctorType:Dissertation
Country:ChinaCandidate:T FangFull Text:PDF
GTID:1102360302989952Subject:Control theory and control engineering
Abstract/Summary:PDF Full Text Request
The modern unmanned aerial vehicle(UAV) played an important role in the military and civil fields, and was being concerned more and more by various countries. Therefore, in order to catch up with and surpass the level of foreigh advanced, it had a very important particle significance and outlook developing interdisciplinary, foundamental, critical , beforehand research on UAV.In recent years, UAVs Coordinated Formation Flight(CFF) was proposed as a new concept to improve the efficiency of UAV completing tasks and widen the scope of its application abroad. The vision-based-sensor UAVs CFF method was from bionic thinking, the rapid development of computer vision methods and hardware and software of processing the video streaming made it possible to achieve the UAVs CFF using visual methods. In fact, the vision-based-sensor has been used to achieve the two-plane formation flight test example.The research on vision-based-sensor of UAVs CFF was related to a number of interdisciplinary . The main study of the paper is as follows considering the key technogies among the field:1)An improved greedy algorithm based active contour model is proposed to realize real-time extraction to the object accurate contour and capture the target feature information of video streaming from the UAV airborne vision-based-sensor. A video image preprocessing method including noise eliminating and stabilization is designed to improve the video image quality; Aiming at the requirements of multiple targets accurate tracking of formation flight video sequences, STK-3D modeling software is utilized to simulate the UAV formation flight. The moving regions are captured by means of double difference operator and multi-resolution connectedness label operator on the simulation video. The improved greedy algorithm is used to extract real-time target contour; the on-line kalman filter group are established to track each target in the video, consequently, the probability way is discussed on how to determine and segment odjects when occlusions occur.2) A way of visual information estimation measurement to relative state of two-plane is proposed using square unscented kalman filter(SUKF). SUKF is introduced to UAVs CFF, and the dynamic model and visual measurement equation of two-plane tracking are established. The process of SUKF applied in visual measurement about relative state of two-plane is described detailedly. The introduction of SUKF insures the positive definiteness of covariance matrix and improves the numerical calculation accuracy of algorithm. The simulation results show their efficiency.3) The traditional PN-OAG is extended to MM-OAG to search the minimum maneuver method of the whole flight track. The continuously updating relative state information from vision-based-sensor is used to judge the obstacle is a collision to object UAV or not. Therefore, The 3D obstacle model is established, the collision criteria is put farward, the design of optimal OAG is introduced. Theoretical analysis and simulation results indicate that MM-OAG has a good prospect.4) CFF physical/virtual prototype united-simulation platform is designed according to existing conditions of laboratory. The system include four subsystems: PP-FCS, UAV dynamic model simulator, VP-FCS, Ground Test System. Each subsystem is described detailedly from designment to implementation, the visual"follower-leader"formation flight simulation test is developed with the united-simulation platform. The correct test results show the correctness of theoretical framework and the feasibility of engineering application of vision-based UAV formation flight. The united-simulation platform provides an important reference and algorithm verification model for CFF simulation study in the future.
Keywords/Search Tags:unmanned aerial vehicle, cooperative formation flight, vision perception, moving object tracking, contour extracting, greedy algorithm, unscented kalman filter, relative state estimation, obstacle avoidance, minimum maneuveobstacle-avoidance guidance
PDF Full Text Request
Related items