The Research Of Flame Detection Based On Video Processing

The state of boiler combustion directly affects the safety and stability of a power plant. The judge of the boiler combustion state is mainly to rely on testing to the state of burner flame in the furnace. The traditional flame detection methods is usually detect the specific light radiation and intensity on flame, which is indirect detection. These methods often can not get the true state of flame and the fire detection result is not satisfactory.Therefor, the paper put forward a video-based image processing method of flame detection. Through acquisition a image of the burner flame, directly judge the combustion flame conditions and make sure the reliability and accuracy of the flame detection. On this foundation, a flame detection system based on programmable Logic Device (FPGA), digital signal processing (DSP) technology has been designed.In the aspect of hardware design, TI corporation 32-bit high speed DSP TMS320F2812 is adopted to serve as the main controller. Apply Altera corporation CycloneII series EP2C5 and High-performance 12-bit A/D converter AD9224 to cooperation with the ahead acquisition chip TC341 to data storage and conversion.In the devise, every functional module was designed, especially design the ahead imaging sensor CCD driver logic, to the effect that the flame image can be accurately acquisition and use the Quartus II software to simulation the timing sequence design. According to characteristic of high speed image acquisition system, the EMI of PCB is also taken into account in order to ensure the performance of the detection system in this paper.The flame image was dispose, before put forward the flame detection algorithm. Because of the disturbance of environment and other factors,the acquisition images of the flame may include a lot of noise signals. Therefore, adopting the multi-image averaging filter and median filter algorithm to reduce image noises.By analising combustion characteristic of the disposed flame image, put forward the flame detecting algorithm based on flame image.