Research On Video Fire Flame Detection System Based On DM8168

Fire disaster is a serious threat to public safety and one of the major disasters in social development. Humans can carry on the reasonable utilization and efficient fire control that is an important symbol of civilization progress. In order to reduce the damage of fire to human and identify the fire flame more quickly and accurately. Needing to find a more efficient way than the traditional flame detection method.At present, scholars at home and abroad use mature computer vision technology and image processing technology to detect the flame, getting a remarkable achievement,but it still have a lot of problems need to slove. This paper will research video fire flame detection system based on this two technologies.In this paper was summarized the principles and characteristics of the video flame detection.Main content has the following several aspects:(1) Using median filtering to de-noising the candidate area image, Using interframe difference method to segment flame region, morphological processing, Comparing the images of the Sobel operator,Roberts operator and Canny operator of these three kinds of edge detection method, discovering of Sobel operator edge detection effect is better.(2) Thus further ensure the flame characteristic data are effective. Then extract the typical characteristics of fire that are the flame color features, flicker frequency characteristics, the circular degree and Angle characteristics. Extracting four characteristic data for the next step of pattern recognition to provide sample data.(3) Using pattern recognition algorithm based on support vector machine (SVM) data train a recognition model through the flame characteristic. In order to get a accurate model of the flame to improve the accuracy of flame recognition and reduce the rate of false positives. This paper will use genetic algorithm (GA), particle swarm optimization (PSO) algorithm and Artificial Bee Colony (ABC) algorithm to optimize the penalty factor c and kernel parameter g of the support vector machine (RBF kernel function). Comparing the advantages and disadvantages of these methods, choose the algorithm with high recognition accuracy as the optimization algorithm of this paper.(4) Designing the fire flame detection system based on TMS320DM8168.Based on the software development environment of DM8168.Building the video input, processing, output, data link.transplantation the main characteristics of flame recognition the algorithmin the DM8168, realization the video detection of fire experiment.