Design And Implementation Of Target Color Recognition And Tracking System Based On FPGA

Target recognition and tracking technology in the visual navigation, intelligent monitoring, Industrial detection and other fields has a wide range of applications. It is one of the important research directions in computer vision. In many areas, target recognition and tracking requires for a good real-time. FPGA has a high throughput capacity of the image data because of parallel processing, and can improve system operating frequency by the way of the pipeline. Along with more integrated, FPGA is becoming to have a smaller volume. The occasion is very suitable for FPGA, which requires real-time to handle large amounts of data and has required volume.Combined the application fields, we have developed the functional requirements of the system. According to the system needs to choose processing platform and tracking algorithm, we chose the FPGA platform and tracking algorithm based on the color to design the identification and tracking system. In the finial, we designed the system overall algorithm flow. We selected the hardware modules according to the application situation, and built the corresponding hardware platform.The color recognition algorithm part involves a lot of image preprocessing algorithms. We have designed and implemented algorithms which identification process involves on the FPGA platform and through the simulation to verify the correctness of the algorithm. The color recognition using dynamic threshold and adaptive color matching template makes the system is robust to changing illumination environment. The using way of the traditional morphological filtering has been improved. The improved algorithm allows the system to select the appropriate filter level according to different sizes of noise. It will maximize to reduce the impact of interference on centroid calculation.The system has used the centroid tracking algorithm. We used centroid calculation method by putting weight to the recognized pixels to reduce the centroid calculation interference by mistakenly identified pixels. In order to improve tracking accuracy, we have transformed the traditional servo. A signal line has been elicited from the position of feedback potentiometer fixedly connected with the rudder plate. Reading the voltage of the signal line by XADC, we can know the current location of the servo. Centroid coordinates and the coordinates of the screen center calculated as the difference, the difference by PID feedback control and the read current position information of the servo are merged. Fusion information generates the appropriate PWM wave, and the system ultimately implements the target object real-time tracking. We have assessed tracking speed and accuracy of the system. In the region of 3 m away from the servo range, track speed ranges up to 23 m/s. It is very fast. The system tracking frequency is servo refresh rate 50 Hz. Tracking accuracy is determined by the maximum precision of track modules, accuracy up to 0.18 °. According to the assessment results, the system meets the design performance requirements.