An improved detection and characterization of active fires and smoke plumes in south-eastern Africa and Madagascar

The present study proposes improved multispectral methods for the detection of vegetation fires and smoke plumes that are applied to south-eastern Africa and Madagascar. Data are provided by the AVHRR sensor onboard the NOAA (11 and 14) satellites. Improvements of a multispectral methods address to fire detection difficulties arising from the low saturation level of AVHRR channel 3, from the presence of clouds and from contrasted vegetation and climate conditions. The methods are based on a multi-channel algorithm using AVHRR data, in visible and thermal ranges. Results are checked against other algorithm and ground concurrent data. It is shown that the presented multispectral methods are able to detect vegetation fires and associated smoke plumes with an improved accuracy. The results evidence clearly the seasonal character of biomass burning. Two maxima are characterized in the reference zone: one in September in Mozambique and the other in October in Eastern Madagascar. We note that fire intensity maxima were accompanied by well developed smoke plumes which could reach more than 50km.     

Detecting and monitoring plumes caused by major industrial accidents with JPLUME, a new software tool for low-resolution image analysis

A new software tool for the automatic detection and monitoring of plumes caused by major industrial accidents is described. This tool has been designed in order to use near real time information as provided by satellite images, perform sophisticated image analysis and elaborate a user-friendly operational environment for the detection of plumes caused by major industrial accidents. The methodology, based on NOAA/AVHRR (Advanced Very High Resolution Radiometer) imagery, uses a two-dimensional feature space in order to discriminate pixels that contain plumes from those that correspond to clouds or to the underlying surface. The two-dimensional feature space is generated by combining AVHRR channels 1 (visible), 2 (near infrared) and 5 (thermal infrared). The software tool proposed has been coded in JAVA2 language, using the concepts of interoperability and object-oriented programming. This study demonstrates the applicability of the tool for the detection of a plume caused by a massive explosion in a firework factory in Enschede, The Netherlands, on May 13, 2000. The effectiveness and reliability of the software tool was found to be satisfactory, as plume was automatically detected and discriminated from the underlying surface.     

雾天条件下基于机器视觉的森林火灾监测

基于机器视觉的森林火灾监测已成为森林火灾监测的一个重要发展方向。烟雾是森林火灾监测的重要指标。然而,诸如云雾和类似烟雾的诸多干扰物降低了火灾识别精度,为此提出了一种结合去云雾和烟雾检测的基于机器视觉的森林火灾监测方法。首先,提取视频中若干帧图像作为样本图像,采用基于Haze-Line的去雾算法对样本图像进行去雾处理。然后,利用基于Horn-Schunck光流法的烟雾检测算法进行烟雾检测,并利用最大类间方差法去除相邻2帧图像间像素质量差异对烟雾检测的影响。最后,利用扩散性分析进行火灾判断。仿真实验及对比分析结果表明,本文方法能够检测出烟雾区域随时间逐渐增加的趋势,从而有效地进行雾天条件下的森林火灾监测,具有更高的准确性和鲁棒性。     

Automatic detection of coastal plumes using Landsat TM/ETM+ images

Riverine fresh water outflows create coastal plumes that are distinguished from surrounding sea water by their specific spectral signature. Coastal waters are unique ecosystems, and they are very important in terms of living resources and oceanographic processes. River plumes and coastal turbid waters have important effects on coastal marine ecosystems, and they also influence marine life cycles, sediment distribution, and pollution. Remote sensing and digital image-processing techniques provide an effective tool to detect and monitor these plume zones over large areas. The primary goal of this study was automatic detection and monitoring of coastal plume zones using multispectral Landsat Thematic Mapper/Enhanced Thematic Mapper Plus (TM/ETM+) imagery. For that purpose, the proposed algorithm exploits spectral features of the multispectral images by using feature extraction and decision-making steps. The procedure has two main stages: (1) some pre-processing operations were applied to the images in order to extract the plume core reflectance values with maximum turbidity and offshore water mass reflectance values; (2) a k-means algorithm was applied with initial seed values of reflectance computed from the pre-processing stage to classify coastal plume zones. Spatial pattern and variability of optical characteristics of coastal plume zones were then defined following the results of the classification process. The algorithm was automatically applied in three different regions with three multispectral Landsat images acquired on different dates, and yielded a very high classification accuracy in detecting coastal plume zones.     

Smoke plume detection in the eastern United States using MODIS

In the eastern United States large amounts of smoke emitted from both wildfires and prescribed fires affect the regional air quality and long‐term climate and may have an impact on public health. Satellite remote sensing is an effective approach for detecting and monitoring the smoke plume. The spectral characteristics of smoke plume are measurably different from those of other cover types, such as vegetation, cloud, snow, and so on. A multi‐threshold method has been developed for detecting smoke plumes with eight MODIS spectral bands based on the analysis of spectral characteristics of different cover types. A series of tests are applied to all pixels in one granule (5‐min measurements) to filter out non‐smoke pixels step by step with water masking. At each step, specific thresholds are utilized. The results have been validated with true color images for a number of cases from different areas and time, showing that the algorithm works well except for a few missing or incorrect identified smoke pixels.     

Evaluation of global fire detection algorithms using simulated AVHRR infrared data

This paper provides a comparison of selected algorithms that have been proposed for global active fire monitoring using data from the NOAA Advanced Very High Resolution Radiometer (AVHRR). A simple theoretical model was used to generate scenes of AVHRR infrared channel 3 and channel 4 data containing fires of various sizes and temperatures in a wide range of terrestrial biomes and climates. Three active fire detection algorithms were applied to the simulated AVHRR images and their performance was characterized in terms of probability of fire detection and false alarm as functions of fire temperature and area, solar and viewing geometry, visibility, season and biome. Additional comparisons were made using AVHRR imagery. Results indicate that while each algorithm has a comparable probability of detecting large (1000m2) fires in most biomes, substantial differences exist in their ability to detect small fires, their tolerance of smoke and neighbouring fires, the number of false alarms, and their overall suitability for global application. An improved automatic algorithm is finally presented. It offers enhanced active fire detection with comparable or reduced false alarm rates in most biomes.     

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.     

Thermal detection of plumes produced by industrial accidents in urban areas based on the presence of the heat island

Detection of plumes produced by industrial accidents using NOAA/AVHRR thermal imagery may be substantially supported in urban areas by the presence of the heat island phenomenon. In this study, an attempt is made to classify the urban web on the basis of the heat island and its impact on the brightness temperatures. Application of the classification scheme on a night-time thermal infrared NOAA-14 image depicting the urban web of Athens demonstrates the potential of this classification for the detection of a plume caused by a fire in a warehouse. Detection of the plume in this case is favoured by the urban heat island phenomenon due to which the urban surface has higher temperature compared to the adjacent environment and the plume above. As a result, distinction of the pixels corresponding to the plume is more effective.     

基于 YOLOX 的小目标烟火检测技术 研究与实现

火灾是日常生活中最常见的社会灾害之一,会对人类的财产、生命安全造成巨大威胁,如何准 确而快速地发现小面积的烟火点并实时发出预警,对维护正常的社会生产具有重要意义。传统的烟火检测算法 通过识别图像的各种低维视觉特征如颜色、纹理等,进而判断烟火的位置,方法的实时性和精度较差。近些年 深度学习在目标检测领域的成就显著,各种基于深度神经网络的烟火检测方法层出不穷,但大部分深度学习模 型在小目标上的检测效果远不及大目标,而烟火检测任务需要在烟火面积很小时就做出及时地识别和预警,才 能避免火势扩大造成更大的经济损失。对此,基于 YOLOX 模型对激活函数和损失函数做出改进并结合数据增 强算法和交叉验证训练方法,实现了更好的小目标检测算法,在烟火检测数据集上获得了 78.36%的 mAP 值, 相比原始模型提升了 4.2%,并获得了更好的小目标检测效果。     

一种结合空间与光谱信息的改进CVA变化检测方法

基于变化向量分析（CVA）的变化检测方法通过直接比较像素差异,能够快速提取多时相影像间的变化信息。尽管如此,由于忽略了像素领域的空间上下文信息及波段之间的差异性和互补性,导致检测结果中难以消除噪声等因素产生的“伪变化”。为此提出了一种结合空间和光谱信息的改进CVA方法。首先,采用主成分分析法对影像进行增强,继而通过构建一种新的多方向差分描述子来提取中心像素的空间上下文信息;在此基础上,提出一种基于相关性的加权融合策略,获得统一的变化强度差分影像;最后,采用EM算法求得变化像素的阈值,继而得到二值检测结果。实验结果表明：所提出的算法能够有效应对“伪变化”的干扰,显著提高变化检测的精度及可靠性。     

一种结合空间与光谱信息的改进CVA变化检测方法

基于变化向量分析（CVA）的变化检测方法通过直接比较像素差异，能够快速提取多时相影像间的变化信息。尽管如此，由于忽略了像素领域的空间上下文信息及波段之间的差异性和互补性，导致检测结果中难以消除噪声等因素产生的“伪变化”。为此提出了一种结合空间和光谱信息的改进CVA方法。首先，采用主成分分析法对影像进行增强，继而通过构建一种新的多方向差分描述子来提取中心像素的空间上下文信息；在此基础上，提出一种基于相关性的加权融合策略，获得统一的变化强度差分影像；最后，采用EM算法求得变化像素的阈值，继而得到二值检测结果。实验结果表明：所提出的算法能够有效应对“伪变化”的干扰，显著提高变化检测的精度及可靠性。     

Satellite-based detection of Canadian boreal forest fires: Development and application of the algorithm

This study presents a comprehensive investigation of fires across the Canadian boreal forest zone by means of satellite-based remote sensing. A firedetection algorithm was designed to monitor fires using daily Advanced Very High Resolution Radiometer (AVHRR) images. It exploits information from multichannel AVHRR measurements to determine the locations of fires on satellite pixels of about 1 km2 under clear sky or thin smoke cloud conditions. Daily fire maps were obtained showing most of the active fires across Canada (except those obscured by thick clouds). This was achieved by first compositing AVHRR scenes acquired over Canada on a given day and then applying the fire-detection algorithm. For the fire seasons of 1994-1998, about 800 NOAA/AVHRR daily mosaics were processed. The results provide valuable nation-wide information on fire activities in terms of their locations, burned area, starting and ending dates, as well as development. The total burned area as detected by satellite across Canada is estimated to be approximately 3.9, 4.9, 1.3, 0.4 and 2.4 million hectares in 1994, 1995, 1996, 1997 and 1998, respectively. The peak month of burning varies considerably from one year to another between June and August, as does the spatial distribution of fires. In general, conifer forests appear to be more vulnerable to burning and fires tend to grow larger than in deciduous forests.     

先验终端像元库支持下的GF-4多光谱影像自动云检测

高分四号卫星是我国第一颗地球同步轨道遥感卫星,以其高频、宽幅的特点,可为我国农业、林业、减灾、气象、环保和水利等应用提供快速、稳定的光学遥感影像,高效的影像自动云检测有助于进一步提高高分四号影像的利用效率。CDAG（Cloud Detection Algorithm-Generating）是一种基于像元组分光谱分析的自动云检测算法,能有效降低混合像元、复杂表面结构和大气等因素的影响。为了探索CDAG算法对于高分4号多光谱影像（GF4-PMS）的云检测应用能力,首先,从高光谱影像（AVIRIS）上选取不同的云类型和各种地表覆盖类型,建立云像元库和地物像元库;其次,基于高光谱像元库和GF4-PMS传感器光谱响应函数模拟出多光谱影像像元库;然后,根据碎云、薄云、厚云及非云像元的光谱差异性分析,将GF4-PMS影像的待检测像元与终端像元进行相似概率分析,实现基于最佳阈值自动迭代的GF4-PMS影像云检测;最后,从云像元正确率、晴空像元正确率、误判率、漏判率等多个指标进行云检测精度验证。结果表明：AVIRIS影像可以有效提取适用于GF4-PMS影像云检测的终端像元库,基于CDAG算法能较好地识别GF4-PMS影像上各种类型的云,对于不同时相、不同下垫面的碎云、薄云、厚云的检测精度可达90%以上。因此,基于先验终端像元库的云检测法对于提升GF4-PMS影像的利用效率具有较好的应用价值。     

Satellite mapping of CO emission from forest fires in Northwest America using MOPITT measurements

We present a study on MOPITT (Measurements Of Pollution In The Troposphere) detection of CO emission from large forest fires in the year 2000 in the northwest United States. Fire data used are from the space-borne Advanced Very High Resolution Radiometer (AVHRR) at 1-km resolution. The study shows that MOPITT can reliably detect CO plumes from forest fires whenever there are >30 AVHRR hotspots in a 0.25° × 0.25° grid, which is comparable to the pixel area of MOPITT in the region. The spatial CO pattern during the fire events is found to be consistent with the location and density of AVHRR hotspots and wind direction. While the increase of CO abundance inside the study area is closely correlated to the AVHRR-derived hotspot number in general (R > 0.75), the non-linearity of fire emission with fuel consumption is also observed. MOPITT can also capture the temporal variation in CO emission from forest fires through 3-day composites so it may offer an opportunity to enhance our knowledge of temporal fire emission over large areas. The CO emission is quantitatively estimated with a one-box model. The result is compared with a bottom-up approach using surface data including burnt area, biomass density, and fire emission factors. If mean emission factors for the region are used, the bottom-up approach results in total emission estimates which are 10%-50% lower than the MOPITT-based estimate. In spite of the limitations and uncertainties addressed in this study, MOPITT data may provide a useful constraint on uncertain ground-based fire emission estimates.     

Automatic Cloud Detection of GF4 Multispectral Imagery Supported by the Priori Terminal Pixel Library

The GaoFen4 (GF4) satellite is China’s first geo-synchronous orbit remote sensing satellite. With the advantages of high frequency and wide width, it can provide fast and stable optical remote sensing images for agricultural, forestry, disaster reduction, meteorology, environmental protection, water conservancy and other applications in China. Efficient image automatic cloud detection helps to further improve the utilization efficiency of GaoFen4 images. CDAG（Cloud Detection Algorihtm-Generating）Cloud detection is an automatic cloud detection algorithm based on spectral analysis of pixel components, which can effectively reduce the influence of mixed pixels, complex surface structure and atmosphere. This paper aims to explore the application of CDAG algorithm in cloud detection of GaoFen4 multispectral imagery (GF4-PMS). Firstly, different cloud types and surface cover types were selected from hyperspectral images (AVIRIS) to establish cloud pixel library and clear sky pixel library. Next, the pixel library of multispectral images was simulated based on Hyperspectral pixel library and spectral response function of GF4-PMS sensor. Then, according to the spectral difference analysis of broken cloud, thin cloud, thick cloud and non-cloud pixel, the similarity probability analysis was performed on the to-be-detected pixel of the GF4-PMS image and the terminal pixel, and the GF4-PMS image cloud detection based on the optimal threshold automatic iteration was realized. Finally, cloud detection accuracy verification was carried out from multiple indicators such as cloud pixel correct rate, clear sky pixel correct rate, false positive rate and missed rate. The results show that AVIRIS images can effectively extract terminal pixel libraries for GF4-PMS image cloud detection. Clouds of Various types on GF4-PMS images can be better identified based on the CDAG algorithm. The accuracy of detection results for broken clouds, thin clouds and thick clouds with different time phases and different underlying surfaces can reach more than 90%. Therefore, the cloud detection method based on the priori terminal pixel library has a good application value for improving the utilization efficiency of GF4-PMS images.     

Technical note: Using NOAA and FY imagery to track plumes caused by the 2003 bombing of Baghdad

In this study, the usability of a 1.1 km satellite receiving station in Greece as an operational tool for monitoring wide scale man-made disasters is demonstrated. The plumes caused by major fires in Baghdad during the 2003 bombing campaign were detected and monitored using the real-time acquisition of National Oceanic and Atmospheric Administration (NOAA) and Feng Yun satellite imagery at the Foundation for Research and Technology - Hellas (FORTH) satellite receiving station. A set of false colour composites of Advanced Very High Resolution Radiometer (AVHRR) and Multichannel Visible and Infrared Scan Radiometer (MVISR) images was used to investigate the dispersion of the plumes, which may contain toxic substances, over the wider Baghdad area, as well as to estimate plume spatial characteristics. In some cases, the area covered by these plumes was estimated to be more than 6000 km².     

结合注意力机制的火灾检测算法

鉴于现有的火灾检测手段大多依赖于感温探测器和感烟探测器,但感温探测器和感烟探测器的探测具有一定的滞后性,无法实时准确地检测出初期火灾的问题,因此,构建了一个大规模多场景的火灾图像数据集;同时对图像数据集进行了火焰和烟雾目标标注,并提出了一种具有注意力机制的火灾检测算法,采用颜色分析的方法检测出图像中火焰和烟雾的疑似区域;再对火焰和烟雾目标的疑似区域进行关注,通过结合深度网络的特征提取能力,得到火灾目标的检测模型;实验结果表明,此方法在检测火灾任务上取得了更优的效果,相比于基于YOLOv3的火灾检测模型,mAP(mean average precision)提高了5.9%,同时满足了实时检测的需求。     

基于图像分割的森林火灾早期烟检测算法研究

针对森林这样的大空间、复杂场景下的火灾检测,提出一种在单帧视频序列图像中的烟检测方法,并研究一种新的超像素合并算法,改进现有的天地线检测算法。该方法对图像进行SLIC（Simple Linear Iterative Clustering）超像素分割,并用一种新的超像素合并算法解决过分割问题;通过改进的天地线分割算法,排除天空中云对于烟检测的干扰;根据光谱特征,运用支持向量机（SVM）对超像素块进行分类。实验结果表明,超像素合并算法高效简洁,易于编程实现,基于图像分割的烟检测技术能排除云雾等噪声对烟雾检测的干扰,在森林场景下的烟雾检测正确率为77%,可以作为人工森林火灾监测的辅助手段。