The Research On Forest Fire Spot And Burn Scar Extraction Algorithms Using HJ-IRS Image

Forest fire not only burn down the forest, destroy the natural environment and ecological balance, but also cause a great threat to human life and property safety. The traditional forest fire monitoring methods only monitor small region and have the low efficiency in obtaining fire information. The remote sensing technology provides a new approach for forest fire monitoring, but the shortcomings of weather satellites such as low spatial resolution and low saturation temperature of the mid-infrared, directly affect the performance of fire spot product.HJ-1B satellite which was launched in2008is part of China’s Small Satellite Constellation for Disaster Prediction and Monitoring Planning. The IRS sensor equipped on HJ-1B satellite has the finer spatial resolution, and its MIR high saturation temperature has reached500K. Based on the HJ-IRS image, the paper presents a new improved forest fire hotspot detection method which is adaptive to HJ-IRS through improving a new cloud and water mask method and adjusting the threshold value in the MODIS version4contextual algorithm. Based on modified NBR index map, a new eight-adjacent image diffusion algorithm which put hotspot generated in last phase as initial seeds is proposed to extract the burn scar area.With Heilongjiang area of China and the Far-east area of Russia as the research object, we prove the algorithms of forest fire spot detection and burn scar area extraction, the main research work as follows:(1) A cloud and water extraction method based on HJ-IRS image was presented by analyzing the characteristics of HJ-IRS sensor.(2) By improving a new cloud and water mask method, adjusting the threshold value in the MODIS version4contextual algorithm, we presented a new improved forest fire hotspot detection algorithm which was adaptive to HJ-IRS, and the algorithm was implemented using IDL(3) Through comparing to the MODIS hotspot products of same area and same time adopting the V4Algorithm, we have demonstrated the accuracy of HJ-IRS fire spot detection algorithm and the pros and cons in low temperature small hotspot detection and geometric location. (4) A modified NBR built by SWIR and LWIR was proposed to reduce the smoke interference, which can enhance the distinction between burn scar area and non burn scar area in the index image.(5) A burn scar area extraction algorithm based on fire spot and modified NBR was presented, which put hotspot generated in last phase as initial seeds and use eight-adjacent image diffusion algorithm to extract the burn scar area.(6) Putting HJ-CCD false color composite image and ETM burn scar product as references, we had demonstrated the accuracy and problem in the algorithm.