An automatic statistical segmentation algorithm for extraction of fire and smoke regions

Estimation of the extent and spread of wildland fires is an important application of high spatial resolution multispectral images. This work addresses a fuzzy segmentation algorithm to map fire extent, active fire front, hot burn scar, and smoke regions based on a statistical model. The fuzzy results are useful data sources for integrated fire behavior and propagation models built using Dynamic Data Driven Applications Systems (DDDAS) concepts that use data assimilation techniques which require error estimates or probabilities for the data parameters. The Hidden Markov Random Field (HMRF) model has been used widely in image segmentation, but it is assumed that each pixel has a particular class label belonging to a prescribed finite set. The mixed pixel problem can be addressed by modeling the fuzzy membership process as a continuous Multivariate Gaussian Markov Random Field. Techniques for estimating the class membership and model parameters are discussed. Experimental results obtained by applying this technique to two Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) images show that the proposed methodology is robust with regard to noise and variation in fire characteristics as well as background. The segmentation results of our algorithm are compared with the results of a K-means algorithm, an Expectation Maximization (EM) algorithm (which is very similar to the Fuzzy C-Means Clustering algorithm with entropy regularization), and an MRF-MAP algorithm. Our fuzzy algorithm achieves more consistent segmentation results than the comparison algorithms for these test images with the added advantage of simultaneously providing a proportion or error map needed for the data assimilation problem.     

Use of NOAA-AVHRR NDVI images for the estimation of dynamic fire risk in Mediterranean areas

Wildfires are a major cause of land degradation in the Mediterranean region due to their frequent recurrence in the same areas. The evaluation of fire risk is therefore of high practical importance, particularly during the summer arid season, when fires are most frequent and harmful. Recent studies have demonstrated that the evaluation of dynamic fire risk can be carried out by the use of remotely sensed images, and specifically of NOAA-AVHRR Normalized Difference Vegetation Index (NDVI) data. This use relies on the sensitivity of the index to vegetation dryness, which is a major predisposing factor for fire occurrence. Several problems, however, remain linked to the spatial variability of the risk in environmentally heterogeneous areas, which requires the application of suitable processing techniques to the low-resolution imagery.The current work reports on the development and testing of different methodologies for estimating dynamic fire risk by the use of NOAA-AVHRR data. The investigation was conducted in Tuscany (Central Italy) using a large archive of fires that occurred in the region and NOAA-AVHRR NDVI data of 16 years (1985-2000). Relying on previous methodological achievements of our group and other research groups, several procedures were tested to extract information related to fire risk from the remotely sensed images. These trials led to define an optimum method which is based on the identification of pixels where the accordance between interyear variations in fire probabilities and NDVI values is maximum. The accuracy of the risk estimates from this optimum method was finally evaluated by a leave-one-out cross-validation strategy. In this way, the performance of the methodology was assessed, together with its potential for operational fire risk monitoring and forecasting in Mediterranean areas.     

Evaluating remotely sensed live fuel moisture estimations for fire behavior predictions in Georgia, USA

Research has shown that remote sensing techniques can be used for assessing live fuel moisture content (LFMC) from space. The need for dynamic monitoring of the fire risk environment favors the use of fast, site-specific, empirical models for assessing local vegetation moisture status, albeit with some uncertainties. These uncertainties may affect the accuracy of decisions made by fire managers using remote sensing derived LFMC. Consequently, the analysis of these LFMC retrieval uncertainties and their impact on applications, such as fire spread prediction, is needed to ensure the informed use of remote sensing derived LFMC measurements by fire managers. The Okefenokee National Wildlife Refuge, one of the most fire-prone regions in the southeastern United States was chosen as our study area. Our study estimates the uncertainties associated with empirical site specific retrievals using NDWI (Normalized Difference Water Index; (R_0.86 − R_1.24) / (R_0.86 + R_1.24)) and NDII (Normalized Difference Infrared Index; (R_0.86 − R_1.64) / (R_0.86 + R_1.64)) that are simulated by coupled leaf and canopy radiative transfer models. In order to support the findings from those simulations, a second approach estimates uncertainties using actual MODIS derived indices over Georgia Forestry Commission stations that provide NFDRS model estimates of LFMC. Finally, we used the FARSITE surface fire behavior model to examine the sensitivity of fire spread rates to live fuel moisture content for the NFDRS high pocosin and southern rough fuel models found in Okefenokee. This allowed us to evaluate the effectiveness of satellite based LFMC estimations for use in fire behavior predictions. Sensitivity to LFMC (measured as percentage of moisture weight per unit dry weight of fuel) was analyzed in terms of no-wind no-slope spread rates as well as normalized spread rates. Normalized spread rates, defined as the ratio of spread rate at a particular LFMC to the spread rate at LFMC of 125 under similar conditions, were used in order to make the results adaptable to any wind-slope conditions. Our results show that NDWI has a stronger linear relationship to LFMC than NDII, and can consequently estimate LFMC with lesser uncertainty. Uncertainty analysis shows that 66% of NDWI based LFMC retrievals over non-sparsely vegetated regions are expected to have errors less than 32, while 90% of retrievals should be within an error margin of 56. In pocosin fuel models, under low LFMC conditions (< 100), retrieval errors could lead to normalized spread rate errors of 6.5 which may be equivalent to an error of 47 m/h in no-wind no-slope conditions. For southern rough fuel models, when LFMC < 175, LFMC retrieval errors could amount to normalized spread rate errors of 0.6 or an equivalent error of 9.3 m/h in no-wind no-slope conditions. These spread rate error estimates represent approximately the upper bound of errors resulting from uncertainties in empirical retrievals of LFMC over forested regions.     

雪盖卫星遥感信息的提取方法探讨

着重论述了从卫星遥感资料中提取雪盖信息的一些方法,结果表明,利用积雪阈值参数从ＮＯＡＡ／ＡＶＨＲＲ图象中提取雪盖信息方法和利用积雪指数（ＮＤＳＩ）从陆地卫星ＴＭ图象中提取雪盖面积的技术,以及利用ＮＯＡＡ／ＡＶＨＲＲ和ＴＭ信息复合的技术,可提高信息获取的精度,具有实用价值。     

Evaluation of a useful method to identify snow‐covered areas under vegetation – comparisons among a newly proposed snow index,normalized difference snow index,and visible reflectance

Objective methods of monitoring snow‐covered areas by optical remote sensing were evaluated using synchronous observations conducted with the passage of the Landsat‐7 satellite over the plains of Niigata prefecture, one of the snowiest regions in Japan. The observations were conducted in the springs of 2002 and 2003. Snow‐covered areas were identified using three methods: (1) visible (red) reflectance, (2) Normalized Difference Snow Index (NDSI) which uses visible and shortwave‐infrared reflectances, and (3) a newly proposed snow index called S3 which uses visible, near‐infrared and shortwave‐infrared reflectances. The Snow‐Cover Ratio (SCR) was defined as the ratio of the number of pixels in snow‐covered areas to the total number of pixels in an image. The threshold value for the three indices used to identify snow‐covered areas was defined as 50% of SCR, which converged to nearly the same value regardless of the images analysed. Under clear conditions, visible (red) reflectance can identify snow‐covered areas accurately if no vegetation is present. NDSI can distinguish snow‐covered areas from mixels (mixed pixels) of snow and vegetation by referring to the Normalized Difference Vegetation Index (NDVI). S3 can distinguish snow‐covered areas from mixels of snow and vegetation without any reference data. S3 is, therefore, more useful than NDSI because it automatically distinguishes snow‐covered areas from mixels of snow and vegetation.     

利用EOS/MODIS资料监测森林火情

分析了MODIS适于火情监测的各通道特性,并基于今年的MODIS资料,运用红外通道及三通道合成的方法,成功地探测到今年发生在我国大兴安岭及境外的多起火情。结果表明,利用红外通道探测法时,将21通道作为红外辐射探测火情的通道能够得到较好的效果。利用三通道合成法时,火灾区域与周边的地物颜色差异较大,可以快速地发现火情,并估计火情的发展趋势。     

基于森林火灾平台的红外图像识别技术研究

森林火灾探测一直是森林资源保护工作中的老大难问题。早期的火灾探测技术多是基于火灾的烟雾和温度特征的,但其判别标准过于单一,误报率较高。随着计算机技术的发展和红外探测器的发明,出现了基于红外图像处理的森林火灾防护技术,该技术主要对红外探测器获得的森林现场的红外图像进行一定的处理后识别图像中是否有火灾出现。本文的主要工作是进行红外图像增强算法的研究,通过对各种算法的描述和仿真实验结果分析,提出一种相对较适合森林背景红外图像的处理算法一将基于频域处理的提升小波变换与直方图修正结合起来的红外图像增强算法。     

Effects of spatial and temporal variation in environmental conditions on simulation of wildfire spread

Environmental conditions, such as fuel load and moisture levels, can influence the behaviour of wildfires. These factors are subject to natural small-scale variation which is usually spatially or temporally averaged for modelling fire propagation. The effect of including this variation in propagation models has not previously been fully examined or quantified. We investigate the effects of incorporating three types of variation on the shape and rate of propagation of a fire perimeter: variation in combustion conditions, wind direction and wind speed. We find that increasing the variation of combustion condition decreases the overall rate of propagation. An analytical model, based on the harmonic mean, is presented to explain this behaviour. Variation in wind direction is found to cause the development of rounded flanks due to cumulative chance of outward fluctuations at the sides of the perimeter. Our findings may be used to develop improved models for fire spread prediction.     

基于PCNN与区域特征的红外与可见光图像融合

针对红外图像与可见光图像融合中容易产生红外目标不明显、对比度不高的问题,提出了一种新的融合算法。该算法创新地将PCNN与区域特征应用到NSCT域内低频和带通子带系数的选择上。通过NSCT分解得到待融合图像的子带系数。运用PCNN对分解后的子带系数进行处理,得到子带系数的点火映射图。低频子带点火映射图采取基于区域标准差的方法选取融合系数。带通子带点火映射图采取基于区域能量的方法选取融合系数。融合图像通过NSCT逆变换可以得到。仿真实验表明与其他算法相比,该算法能够得到红外目标突出、质量更好的融合图像,图像客观评价指标提升明显。     

Effect of scan angle on AVHRR fire detection accuracy in interior Alaska

Some AVHRR fire detection studies have excluded pixels that exceeded an arbitrary scan angle. This exclusion seems to be based on the distortion of pixels at high scan angles and the well-documented effects of scan angle on the Normalized Difference Vegetation Index. However, excluding high scan angle pixels reduces the temporal resolution of an AVHRR fire detection system, especially at high latitudes. High scan angle pixels may be less obscured by smoke compared to near-nadir pixels. We tested the effect of scan angle on AVHRR fire detection by comparing scan angle classes of less than and greater than 25 from sixteen AVHRR images of interior Alaska. At scan angles under 25, 35/86 (40%) of fire observations were detected. At scan angles over 25, 24/53 (45%) of fire observations were detected. For interior Alaska, where cloudy, low-fuel conditions exist, and 8-12 AVHRR images are available daily, we recommend not excluding high scan angle pixels for wildfire detection.     

基于NDTI指数的MODIS火情监测研究

在分析燃烧热点光谱辐射特征的基础上,提出基于MODIS红外辐射的NDTI指数火情监测模型,给出了该监测模型的详细算法,并通过几起较大森林火灾对该模型进行了分析和验证。这种双通道比值指数和NDVI指数一样具有衰减仪器噪声和抵消常规大气影响的优点。MWIR 4μm火点辐射通道和LWIR 11μm背景分量通道的组合突出了火点信息。研究表明,选择合适的NDTI门限,NDTI指数能够准确检测火点。MODIS NDTI火情指数模型可以推广到同类探测器的应用中,例如NOAA AVHRR、GOES VAS以及即将运行的NPP VIIRS和中国风云3号FY-3平台,特别是MODIS的下一代多光谱探测器NPOESS NPP VIIRS设计继承了MODIS的优势,为NDTI算法模型提供了广阔的应用前景。     

基于嗅觉神经网络仿生模型的纹理图像生成算法

提出了一种基于嗅觉系统生成纹理图像的仿生模型。该模型结构模拟嗅觉神经网络的结构,利用Logsitic函数的混沌特性调整每次迭代过程中的模型参数,使用简单的周期函数作为模型节点的激活函数实现纹理的重复,并引入随机噪声来模拟脑在进行信息处理时的背景噪声。实验结果表明,该模型可以生成丰富而多变的纹理图像,引入的随机噪声也起到了积极的作用,可以明显地丰富纹理图像的变化。此外,模型生成纹理图像的效率也高于传统的BP神经网络模型。     

Monitoring herbaceous fuel moisture content with SPOT VEGETATION time-series for fire risk prediction in savanna ecosystems

This paper evaluated the capacity of SPOT VEGETATION time-series to monitor herbaceous fuel moisture content (FMC) in order to improve fire risk assessment in the savanna ecosystem of Kruger National Park in South Africa. In situ herbaceous FMC data were used to assess the Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), Vegetation Dryness Index (VDI), Improved VDI (IVDI), and Accumulated Relative NDVI Decrement (ARND) during the dry season. The effect of increasing amounts of dead vegetation on the monitoring capacity of derived indices was studied by sampling mixed live and dead FMC. The IVDI was proposed as an improvement of the VDI to monitor herbaceous FMC during the dry season. The IVDI is derived by replacing NDVI with the integrated Relative Vegetation Index (iRVI), as an approximation of yearly herbaceous biomass, when analyzing the 2-dimensional space with NDWI. It was shown that the iRVI offered more information than the NDVI in combination with NDWI to monitor FMC. The VDI and IVDI exhibited a significant relation to FMC with R² of 0.25 and 0.73, respectively. The NDWI, however, correlated best with FMC (R² = 0.75), while the correlation of ARND and FMC was weaker (R² = 0.60) than that found for NDVI, NDWI, and IVDI. The use of in situ herbaceous FMC consequently indicated that NDWI is appropriate as spatio-temporal information source of herbaceous FMC variation which can be used to optimize fire risk and behavior assessment for fire management in savanna ecosystems.     

Multi-temporal compositing approaches for SPOT-4 VEGETATION

Five image compositing criteria for Système pour l'Observation de la Terre (SPOT)-4 VEGETATION data were compared. Selecting the pixel that satisfies a given criterion from a monthly set of daily images produces the composite image. The criteria applied are: maximization of Normalized Difference Vegetation Index (NDVI); minimization of the red channel; maximization of NDVI followed by minimization of short-wave infrared (SWIR); selection of the third lowest value of near-infrared (NIR); and selection of the third lowest value of an albedo-like index, designated by 'darkness'. Visual and quantitative analysis indicate that the last approach produces the cleanest images.     

C-V模型在红外视频火灾图像分割中的应用

针对大空间中红外视频火灾图像边缘模糊,不易准确分割问题,研究了一种基于背景差分和C-V模型的分割方法。通过背景差分得到运动图像;利用形态学处理得到完整的运动区域,并获得其外接矩形;以外接矩形作为C-V模型的初始轮廓曲线进行分割,得到封闭、完整的运动目标轮廓。该算法避免了对整幅图像分割,减少了运算量。通过实验仿真并与阈值分割算法比较,证明了该算法的准确性和有效性,有利于下一步火灾特征提取与识别。     

Comparison of the use of WiFS and LISS images to estimate the area burned in a large forest fire

We evaluate the utility of medium spatial resolution images from the Wide Field Sensor (WiFS) for the estimation of the area burned in a large fire. The performance of methodologies using these images is compared with similar methodologies using high spatial resolution image from the Linear Imaging and Self Scanning Sensor (LISS-III) and other ancillary data. Both sensors are located onboard the Indian Remote Sensing Satellite 1C (IRS-1C). The post-fire LISS image was analysed by means of Matched Filtering (MF) techniques. Two WiFS images (pre- and post-fire) were analysed using MF techniques and also by means of changes in the Normalized Difference Vegetation Index (NDVI). Ground data were used to classify the three thematic images obtained in several post-fire classes. The results show a greater proportion of transition areas between burned and unburned places and a slightly larger area burned estimation in the WiFS than in the LISS analysis. Nevertheless, the results obtained, and the comparisons with ground data, indicate that medium spatial resolution images' estimation of the area burned is a useful tool at regional and national scales.     

Wildfire temperature and land cover modeling using hyperspectral data

Wildfire temperature retrieval commonly uses measured radiance from a middle infrared channel and a thermal infrared channel to separate fire emitted radiance from the background emitted radiance. Emitted radiance at shorter wavelengths, including the shortwave infrared, is measurable for objects above a temperature of 500 K. The spectral shape and radiance of thermal emission within the shortwave infrared can be used to retrieve fire temperature. Airborne Visible Infrared Imaging Spectrometer (AVIRIS) data were used to estimate fire properties and background properties for the 2003 Simi Fire in Southern California, USA. A spectral library of emitted radiance endmembers corresponding to a temperature range of 500-1500 K was created using the MODTRAN radiative transfer model. A second spectral library of reflected solar radiance endmembers, corresponding to four vegetation types and two non-vegetated surfaces, was created using image spectra selected by minimum endmember average root mean square error (RMSE). The best fit combination of an emitted radiance endmember and a reflected solar radiance endmember was found for each spectrum in the AVIRIS scene. Spectra were subset to reduce the effects of variable column water vapor and smoke contamination over the fire. The best fit models were used to produce maps of fire temperature, fire fractional area, background land cover, land cover fraction, and RMSE. The highest fire temperatures were found along the fire front, and lower fire temperatures were found behind the fire front. Saturation of shortwave infrared channels limited modeling of the highest fire temperatures. Spectral similarity of land cover endmembers and smoke impacted the accuracy of modeled land cover. Sensitivity analysis of modeled fire temperatures revealed that the range of temperatures modeled within 5% of minimum RMSE was smallest between 750 and 950 K. Hyperspectral modeling of wildfire temperature and fuels has potential application for fire monitoring and modeling.     

红外图像处理技术在苹果早期淤伤检测中的应用

在苹果品质检测中,淤伤是影响检测效率的一个重要原因,针对苹果淤伤检测这一难题,提出应用红外图像处理技术解决.根据红外热图像成像原理及其特点,采用风扇加热与冷却两种实验方法.在实验过程中,红外摄像机拍摄获得苹果红外图像,选用优越的混合图像增强算法等红外图像处理方法识别早期淤伤.实验结果表明,应用红外图像技术可以检测苹果早期淤伤,且效率迭96%以上.     

基于空域自然场景统计的无参考立体图像质量评价模型

针对现有的评价方法大都将图像变换到不同的坐标域问题,提出一种基于空域自然场景统计(NSS)的通用型无参考立体图像质量评价模型。在评价中为了更好地结合人类双目视觉特性, 将左右图像融合成一幅独眼图;评价模型首先统计独眼图归一化亮度(CMSCN)系数分布规律,进而对独眼图提取空域自然场景统计特征;其次,统计视差图归一化亮度(DMSCN)系数的分布规律,并对用光流法得到的视差图提取同样的特征;最后,通过支持向量回归(SVR)建立立体图像特征信息与主观评价值(DMOS)之间的关系,从而预测得到图像质量的客观评价值。实验结果表明,该评价模型对立体数据测试库进行评价,其Pearson线性相关系数(PLCC)和Spearman等级相关系数(SROCC)值均在0.94以上;对于非对称立体图像库,PLCC和SROCC值分别接近0.91和0.93。该模型能够很好地预测人眼对立体图像的主观感知。     

基于“北京一号”小卫星数据的密云水库流域植被覆盖度估算方法比较

“北京一号”小卫星（BJ-1）是一颗拥有高时频、覆盖宽度大等优势的对地观测小卫星。运用BJ-1遥感数据，以密云水库流域为研究区域，通过NDVI像元二分法、三波段梯度差法估算其植被覆盖度，并尝试利用重归一化植被指数（RDVI）法进行植被覆盖度估算。通过对3种方法估算结果的比较发现：RDVI法的估算结果与实测值的吻合度较高，而三波段梯度差法则出现较大误差，NDVI像元二分法的估算结果精度居中。结果表明：运用BJ-1数据，采用RDVI法可以有效地进行连续的、大范围的植被覆盖度估算。