Design And Implementation Of Airborne Real-time Image Tracking System Based On Dual-core Mechanism

Based on the project of Guangdong Provincial Department of Science and Technology, which focuses on the monitoring and controlling fire disaster UAV(Unmanned Aerial Vehicle) at low altitude, this paper built a set of airborne real-time image tracking system. The system major component consists of four section: ground station section?system communication section? airborne PTZ control section and image recognition section, where the image recognition and PTZ control are the core parts. The system, ultimately, realizes the function of real-time data processing and target tracking by using TI dual-core mechanism?the numerical processing capacity of DSP algorithms servers.The system contains mainly two parts, which are the ground software section and airborne software section.The ground software section is the entrance of man-machine interaction, coding in GTK. It realizes the function of dispatching and receiving control instructions and the real-time displays of video. The ground station section and airborne software section exchange data via UDP technology. And a data communication protocol based on the application layer is designed. It is the standard for coding and decoding of information in air and ground communication software. Finally, real-time displays of video is accomplished by the combine with open source VLC media player and the real-time video streaming provided by our webcams.Airborne software section is the key part of the thesis. It is the center of the data processing system which mainly responsible for image recognition and PTZ control. The core technology of this part is the cooperative work between DSP core and ARM core on the DM3730 core processor. The system real-time performance improved dramatically, using the strong numerical computation capability of DSP and dual-core in parallel. Where ARM core is mainly responsible for I/O event processing, including coding and decoding of control signals, video capture, PTZ control and DSP operation control, and it is mainly accomplished through thread technology. V4L2 technology is used for video capture, and ONVIF protocol and PID theory are used for PTZ control. Meanwhile, DSP core is the operating platform of image algorithm. In combination with ARM core, they constitute the C / S service structure. Taking DSP as the image algorithm server, and ARM core as the consumer side, the algorithm is called by way of remote procedure call. Its implementation utilizes TI Codec Engine software framework to complete the communication and data sharing between cores. In this paper, classical camshift algorithm, implemented by TI DSP Algorithm Standard x DAIS, is adopted for image recognition, which has the advantages of small computation burden, real-time, good recognition performance.Finally, a combined adjusting test was made on the entire airborne system which analyzing the ability of anti-interference and speed tracking. Experiment shows that the system is of highly efficiency, fantastic real- time, as well as user-friendly. What's more, the operating speed reaches as fast as 21.6 fps when the input of the video frame is 320*240.