The Design Of Flame Detection System Based On ARM

With the development of computer technology, image processing, image identification technology has been applied in all walks of life. Flame detection is usually used in the field of early fire alarm, furnace flame identification. Most of their sensors are temperature sensor smoke sensor, light sensor or Infrared detectors. Traditional identification methods are powerless to deal with the torch flame identification in such large space with outdoor complex background. Therefore we use digital image processing techniques to identify the torch flame.As power resources are limited on the oil platform in the sea, we use embedded solutions. We use S3C2410A as main processor which is base on ARM920T and whose bit width is 32 and speed is up to 203 MHz.We expand the system's memory to run their software systems. The system use a modular designing method, we have designed a power module,a reset module,a USB port based on which we can use a USB interface CCD to detect image. In order to facilitate system update, we integrated an ethernet interface in it.On the basis of the existing image processing technology, we analysis furnace flame identification technology and the fire alarm technology based on image processing technology, and combining the characteristics of the torch flame, designed a BP neural network Algorithm to identify the flames. System captures an image every 100 millisecond, then turn the color image to gray scale image, using mathematical morphology to do filtering, gray-scale transformation and such as image preprocessing and image enhancement. Then trace the object contour, and calculate its various morphological parameters. Match the same object'contour in different images. Use the change of the same object's morphological parameters as the neural network input parameters, to judge whether the flame in the background.We uses LINUX OS, chose WINDOWS XP+VMWARE+REDHAT 9 VMWARE LINUX as a development platform, use a cross-compiler and visualization of QTOPIA solution as develop tools.