Research On Extraction Of Burned Area Based On Image Enhancement

Fire is a global natural disaster and has a close relationship with global climate change.The burned area is the area where the vegetation has not returned to its original state after burning.Access to information on burned areas rapidly and precisely can contribute to studying the carbon cycle and climate change.In addition,burned area mapping can also be used to protect and restore fire-affected ecosystems.Remote sensing has the advantages of getting ground information rapidly,widely and consistently.It has become an important way of extracting burned areas.In recent years,a variety of burned area mapping products have been produced based on different remote sensing data.These products can provide global information for research on ecosystem,climate and carbon cycle models,and forest fire dynamic.The Medium Resolution Spectral Imager(MERSI)onboard the FY-3C satellite launched in China in 2013 features multi-spectral and high-resolution imaging that can detect the reflection radiation of visible and near-infrared,and the emission radiation of thermal infrared.The data can generate daily global seamless coverage images,observe the earth at a higher frequency,and has certain advantages in global disaster monitoring.The paper uses FY-3C MERSI data,makes full use of the characteristics of the burned area,and studies the extraction of the burned area based on the image enhancement method.Wildfire will destroy the physiological structure of the plant and reduce the vegetation coverage.These two factors change the spectral characteristics of the vegetation on the image.In many remote sensing methods of mapping burned area,the most common method of enhance burned area is building indices with bands which are sensitive to charcoal,ashes,moisture and vegetation changes.According to the band characteristics of FY-3C MERSI data,this paper firstly analyzes and selects the bands and indices that are sensitive to burned area.Based on their features of burned area in the uni-temporal and multi-temporal images,the burned areas in the study areas were initially extracted.Next,we consider that the burned area has saliency features on the entire image.Many methods for identifying the salient features of images have gradually been used in remote sensing image processing.FASA(fast,accurate,and size-aware saliency object detection)is used in this paper to enhance the information of burned area.It extracts salient objects by estimating the location and color characteristics of the salient object,which is both fast and accurate.The method first quantifies the color in a uniform CLEL*a*b color space.Then by calculating the spatial center and variance of the quantified color,the position and size of the saliency object are obtained.A relationship model of the position,size and saliency probability of salient object is constructed,and the salience probability of saliency is computed by estimated value of the position and size of salient objects.Global contrast is calculated with quantized colors,and combined with saliency probability to generate the final saliency map.We consider that based on the preliminary extraction of the burned area using the band and spectral indices,the burned areas in the image can be enhanced by FASA and then can be further processed by morphological methods to achieve the purpose of accurate extraction.Finally,two study areas were selected in this paper to enhance and extract the burned areas using FY-3C MERSI data.Visual interpretations are used to evaluate the experimental results of the proposed method and compared with the results of NBR threshold method.The results show that the kappa coefficient of the saliency enhancement method in the two research areas reaches more than 0.67,0.2 higher than NBR threshold method.We suggest that the image enhancement method for mapping burned area is high,and the influence of vegetation change caused by non-burned has little influence on it,and the method has a certain stability compared with the NBR threshold method.The study areas selected in this paper have similar conditions.In the following study the method can be applied on different types of burned areas to explore the applicability of the methods under different climates and vegetation types.In addition,the performance of the newly launched FY-3D MERSI is higher than that of the FY-3C MERSI.This data can be used for subsequent studies in the future.