| Person recognize is very important in the field of fire control and personnel security. This paper put forward a kind of embedded person recognize system combined the technology of image recognition and target tracking. This system did the image analysis processing, detected the number of people and their own direction, and uploaded the results to display.This system uses FPGA+ARM structure, first collecting image from CMOS digital sensor, and then transmiting the image date to SRAM in the control of FPGA. In the system, FPGA and ARM communicate through SPI bus, and SRAM transmites the image date to ARM in the control of FPGA. ARM will process the image data further, identify people from every frame, judge the direction and number of people by a few successive frames, and then transmite the final calculated results to PC. PC will display the results.In ARM, the system will zoom image to1/9to improve the speed of system; median filter is used to reduce the image noise, and improve the accuracy; eliminates the background in the image with the background difference method, reduces the interference of the background on the target object; processes the gray image with an improved global threshold value method to get the binary image; divides the connected domain of binary image, deletes the connected domain which did not conform to the image size of person; does edge detection to get the outline. After the image preprocessing, the system will identify people from signal image with an improved method, complete the person recognize with tracking people from successive a few frames of the image.The4groups of experiment verifies the accuracy and real-time performance of the system, it shows that when one person through, there will be misjudgment rarely, the accuracy of system can reach above90%; when many people through in the same time, the accuracy of system will decline, can reach above85%. System can finish person recognize successfully under the good light condition. |
PDF Full Text Request