Effective Dynamic Object Detecting for Video-Based Forest Fire Smog Recognition

This paper presents an improved novel dynamic smoke detecting method for automatic forest fire surveillance with long-distance video. The first part describes the improved dynamic object detecting technique, that is, the finite thresholding processing to each differential frame after multi-frame temporal difference operation to extract the persistent dynamic behavior of forest smoke from serial forest fire frames. The second part deals with the special characteristics of a real fire smoke (persistence, increase, for example) to discriminate the similar natural phenomena effectively. The early fire which is even not easy to find by manpower was detected with the method in some Forest Park. At the same time the rate of false alarm also is kept within 15%.