The Research Of Image Motion Detection Technology Based On FPGA

The clarity of video image sequence and the information contained in them influenced people’s mastering and judgment towards all kinds of information to a large extent. However, in the cases of imaging system in a mobile situation, such as the imaging equipment of vehicle, shipborne, spaceborne or hand-held, there is always mechanically dithered ring or random jitter on imaging vector, which will generate the image motion that brings the result of imaging blurring or smearing effect. Image stabilization technology can decrease or even eliminate the phenomenon of image sequence blurring caused by random movement or jitter of imaging equipment. The key to image stabilization technology lies in acquisition of image motion precisely in real time, that is to say, the measurement accuracy of image motion determines the image effect and precision of imaging system.This thesis first introduces the background and significance of the subject as well as investigates the relative research development and trend of image stabilization technology. On this basis, the current hardware platforms for image motion measurement system are introduced. Secondly, it analyses the influences and causes of image motion, studies the performance improvement of image motion algorithm in aspects of algorithm’s precision, robustness, computing efficiency, which lays a solid foundation for the later research of image motion measurement algorithm based on FPGA. The thesis focus on the research of the image motion measurement of three different methods and proposes the corresponding optimized algorithm which is fit for FPGA’s process on the basis of introduction of basic principles for the traditional joint transform correlation method and phase correlation method. It analyses the influences of the sub-image sizes, relative displacement, threshold value and image distortion to the algorithm accuracy of improved image motion measurement. Then, it proposes the measurement algorithm for motion vector based on corner point correlation. Numerous simulations have been carried on in MATLAB environment for the three algorithms, which simulates the process pattern of FPGA. Through measuring, it shows that the RMS error can be controlled within 0.1 pixels. After analyzing the advantages and disadvantages of the three algorithms, the optimized phase correlation method in FPGA is chosen. Then, the hardware circuit for measuring image motion based on FPGA is designed. The main work includes demonstration for system design, overall framework design, electronic components selection, schematic diagram design and detection for the final hardware circuit. Finally, the algorithm is engineered. The system can be divided into parts as grabbing image sequence, intermediate result buffering, detail process of the algorithm and final result output. The core algorithm involves windowing process for image sequence, kinds of mathematic computation, FFT algorithm, magnitude operation and computation of floating point and so on. Verilog HDL is used in system design, synthesis and implementation are executed in ISE which is for programming design of XILINX. The Modelsim is used for simulation of all modules and it is verified on the FPGA processing board.The experiment result shows that the image motion detection system based on FPGA designed in this paper works steadily. The output of the system can reach more than 400 image motions per second. In addition, frame dropping hardly occurs. The measuring accuracy in stationary state can arrive at second decimal place. The processing curve calculated for image motion in dynamic state is consistent with the waveform of signal generator. Furthermore, this thesis looked into future on the basis of the summary of the subject study and pointed out some places to be improved, which may provide a valuable reference for some other designs of image motion measurement system.