A Saliency-Based Method for Early Smoke Detection in Video Sequences

Video-based smoke detection requires suspected smoke regions to be segmented from the complex background in the initial stage of detection. This segmentation is also important to the subsequent processes of detection. This paper proposes a novel method of segmenting a smoke region in smoke pixel classification based on saliency detection. A salient smoke detection model based on color and motion features is used. First, smoke regions are identified by enhancing the smoke color nonlinearly. The enhanced map and motion map are then used to measure saliency. Finally, the motion energy and saliency map are used to estimate the suspected smoke regions. The estimation result is regarded as our final smoke pixel segmentation result. The performance of the proposed algorithm is verified on a set of videos containing smoke. In the experiments, the method achieves average smoke segmentation precision of 93.0%, and the precision is as high as 99.0% for forest fires. The results are compared with those of three other methods used in the literature, revealing the proposed method to have both a better segmentation result and better precision. We also present encouraging results of smoke segmentation in video sequences obtained using the proposed saliency detection method. Furthermore, the proposed smoke segmentation method can be used for real-time fire detection in color video sequences.     

Deep Belief Network For Smoke Detection

Forest fire is an serious hazard in many places around the world. For such threats, video-based smoke detection would be particularly important for early warning because smoke arises in any forest fire and can be seen from a long distance. This paper presents a novel and robust approach for smoke detection that employs Deep Belief Networks. The proposed method is divided into three phases. In the preprocessing phase, the region of high motion is extracted by background subtraction method. During the next phase, smoke pixel intensities are extracted from the Red, Green and Blue and Luminance; Chroma:Blue; Chroma:Red color spaces for foreground regions. Subsequently, second feature which is based on texture is computed for detecting smoke regions in which Local Extrema Co-occurrence Pattern, an improved version of local binary patterns are extracted from different foreground regions which compute not only texture of smoke but also intensity and color of smoke using Hue Saturation Value color space. Finally, Deep Belief Network is employed for classification. The proposed method proves its accuracy and robustness when tested on different varieties of scenarios whether wildfire-smoke video, hill base smoke video, indoor or outdoor smoke videos.     

Erratum to: Video Fire Smoke Detection Using Motion and Color Features

A novel video smoke detection method using both color and motion features is presented. The result of optical flow is assumed to be an approximation of motion field. Background estimation and color-based decision rule are used to determine candidate smoke regions. The Lucas Kanade optical flow algorithm is proposed to calculate the optical flow of candidate regions. And the motion features are calculated from the optical flow results and use to differentiate smoke from some other moving objects. Finally, a back-propagation neural network is used to classify the smoke features from non-fire smoke features. Experiments show that the algorithm is significant for improving the accuracy of video smoke detection and reducing false alarms.     

Video Fire Smoke Detection Using Motion and Color Features

A novel video smoke detection method using both color and motion features is presented. The result of optical flow is assumed to be an approximation of motion field. Background estimation and color-based decision rule are used to determine candidate smoke regions. The Lucas Kanade optical flow algorithm is proposed to calculate the optical flow of candidate regions. And the motion features are calculated from the optical flow results and use to differentiate smoke from some other moving objects. Finally, a back-propagation neural network is used to classify the smoke features from non-fire smoke features. Experiments show that the algorithm is significant for improving the accuracy of video smoke detection and reducing false alarms.     

Video-based smoke detection with histogram sequence of LBP and LBPV pyramids

Video surveillance systems are widely applied in a variety of fields. Hence, video-based smoke detection is regarded as an effective and inexpensive way for fire detection in an open or large spaces. In order to improve the efficiency of the video-based smoke detection, a novel video-based smoke detection method is proposed by using a histogram sequence of pyramids. The method involves four steps. Firstly, through multi-scale analysis, a 3-level image pyramid is constructed. Secondly, local binary patterns (LBP), which are insensitive to image rotation and illumination conditions, are extracted at each level of the image pyramid with uniform pattern, rotation invariance pattern and rotation invariance uniform pattern to generate an LBP pyramid. Thirdly, local binary patterns based on variance (LBPV) with the same patterns are also adopted in the same way to generate an LBPV pyramid. And fourthly, histograms of the LBP and LBPV pyramids are computed, and then all the histograms are concatenated into an enhanced feature vector. A neural network classifier is trained and used for discrimination of smoke and non-smoke objects. Experimental results show that the features are insensitive to rotation and illumination, and that the method is feasible and effective for video-based smoke detection at interactive frame rates.     

多传感器融合的火灾识别算法研究

针对传统火灾探测存在反应滞后,误报率高等问题,提出一种基于多传感器融合的火灾识别算法。采用运动和颜色检测提取可见光图像的疑似火灾区域,同时采用阈值分割提取红外图像的疑似火灾区域。将两者疑似火灾区域分别进行特征提取,再根据所获特征进行基于 SVM 的融合火灾识别,通过传感器检测实时的温度与烟雾浓度,从而实现可靠和快速的火灾识别。实验结果表明,该火灾识别算法在可接受的时间范围内准确率高且鲁棒性强。     

输电线路山火无人机监测与灭火技术研究

应用无人机搭载三维激光,识别输电线路、杆塔及下方植被,进行山火风险评估。无人机搭载YCbCr颜色空间检测分析模块进行山火识别。提出了无人机弹道发射、投掷灭火球以及小流量防蒸发喷射等输电线路初发山火的灭火方法,可快速扑灭初发山火。     

Fire detection and 3D surface reconstruction based on stereoscopic pictures and probabilistic fuzzy logic

In this paper, a new fire detection method is proposed, which is based on using a stereo camera to calculate the distance between the camera and the fire region and to reconstruct the 3D surface of the fire front. For the purpose of fire detection, candidate fire regions are identified using generic color models and a simple background difference model. Gaussian membership functions (GMFs) for the shape, size, and motion variation of the fire are then generated, because fire regions in successive frames change constantly. These three GMFs are then applied to fuzzy logic for real-time fire verification. After segmentation of the fire regions from left and right images, feature points are extracted using a matching algorithm and their disparities are computed for distance estimation and 3D surface reconstruction. Our proposed algorithm was successfully applied to a fire video dataset and its detection performance was shown to be better than that of other methods. In addition, the distance estimation method yielded reasonable results when the fire was a short distance from the camera and the reconstruction of the 3D surface showed a shape that was almost the same as that of the real fire.     

三光集成型精确山火定位监测仪

描述了该山火监测设备的硬件组成、火光和烟雾自动识别算法和高精度定位方法,通过真型电力线路故障引发山火的试验和历史山火烟雾图像验证了该设备的指标和性能。实际试验表明,该设备在传统的可见光红外监测摄像头的基础上,增加火光和烟雾自动识别算法和高精度定位方法,能够在山火发生初始阶段有效监测识别,对于电力线路防山火有较大实际意义。     

有雾天气下输电线路山火烟雾检测方法研究

输电线路多处于环境复杂的山林中,早期山火发生时经常以烟雾的形式呈现,而在有雾状况下的山火烟雾检测方法的研究却很少见。针对有雾天气状况时的山火检测,提出一种去雾图像增强方法,首先对图像局部均衡化处理,再对全局利用改进的单尺度Retinex 方法做增强处理,并使用基于卷积神经网络的山火烟雾检测网络来检测早期山火发生时产生的烟雾。实验结果表明,基于局部和全局的图像增强方法可使山火烟雾检测准确率有明显提升,通过卷积神经网络的烟雾检测准确率达到97.2%。     

An Improved Probabilistic Approach for Fire Detection in Videos

This paper proposes an improved probabilistic approach using two improved feature representations. These features are color and motion. First, an improved probabilistic model for color-based fire detection is proposed, and candidate fire regions are generated from this model. Then, an improved motion feature is used for final decision. The performance of the proposed approach showed about 0.2758 accuracy in false positive rate, and 0.2636 accuracy in false negative rate on a benchmark fire video database, which represents a decrease of 46.6% in false positive rate, and a decrease of 52.1% in false negative rate compared to the probabilistic approach.     

Video based wildfire detection at night

There has been an increasing interest in the study of video based fire detection algorithms as video based surveillance systems become widely available for indoor and outdoor monitoring applications. A novel method explicitly developed for video based detection of wildfires at night (in the dark) is presented in this paper. The method comprises four sub-algorithms: (i) slow moving video object detection, (ii) bright region detection, (iii) detection of objects exhibiting periodic motion, and (iv) a sub-algorithm interpreting the motion of moving regions in video. Each of these sub-algorithms characterizes an aspect of fire captured at night by a visible range PTZ camera. Individual decisions of the sub-algorithms are combined together using a least-mean-square (LMS) based decision fusion approach, and fire/nofire decision is reached by an active learning method.     

交通视频图像的多尺度自适应阈值边缘检测

基于视频图像的交通流参数检测,车辆检测是其关键。车辆检测常用的有帧差法、背景差法和边缘检测法等。边缘检测是适应性最好的方法。本文把基于小波变换的多尺度自适应阈值图像边缘检测的新方法应用于交通图像边缘检测。通过计算机仿真对本算法进行了验证,结果明显好于采用其它几种经典方法的边缘检测。     

Multi-point vibration measurement and mode magnification of civil structures using video-based motion processing

Image-based vibration measurement has gained increased attentions in civil and construction communities. A recent video-based motion magnification method was developed to measure and visualize small structure motions. This new approach presents a potential for low-cost vibration measurement and mode shape identification. Pilot studies using this approach on simple rigid body structures were reported. Its validity on complex outdoor structures has not been investigated. In this study, a non-contact video-based approach for multi-point vibration measurement and mode magnification is introduced. The proposed approach can output a full-field vibration map that increases the efficiency of the current structural health monitoring (SHM) practice. The multi-point approach is developed based on the local phases which also fill the gap of the existing intensity-based multi-point vibration measurement. As an extension of the phase-based motion magnification, the multi-point measurement result is then integrated with the maximum likelihood estimation (MLE) to estimate the magnified frequency bands at each identified structure mode for operational deflection shape (ODS) visualization. This proposed method was tested in both indoor and outdoor environments for validation. The results show that using the developed method, mode frequencies and mode shapes of multiple points in complex structures can be simultaneously measured. And vibrations in each mode can be visualized separately after magnification.     

An effective four-stage smoke-detection algorithm using video images for early fire-alarm systems

This paper proposes an effective, four-stage smoke-detection algorithm using video images. In the first stage, an approximate median method is used to segment moving regions in a video frame. In the second stage, a fuzzy c-means (FCM) method is used to cluster candidate smoke regions from these moving regions. In the third phase, a parameter extraction method is used to extract a set of parameters from spatial and temporal characteristics of the candidate smoke regions; these parameters include the motion vector, surface roughness and area randomness of smoke. In the fourth stage, the parameters extracted from the third stage are used as input feature vectors to train a support vector machine (SVM) classifier, which is then used by the smoke alarm to make a decision. Experimental results show that the proposed four-stage smoke-detection algorithm outperforms conventional smoke-detection algorithms in terms of accuracy of smoke detection, providing a low false-alarm rate and high reliability in open and large spaces.     

A smoke model and its application for smoke management in an underground mass transit station

This paper discusses the development of a smoke model for CFD. The model evaluates smoke visibility based on line of sight. Using a compartment fire, the deficiency in determining visibility by the conventional surface-based approach is first demonstrated. Smoke management in an underground rail station is investigated using the smoke model. For a medium growth rate fire, the results show that the platform is blocked by smoke within 2–3 min. On the mezzanine, the designed smoke exhaust controls the smoke only for a limited time of less than 4 min. The variations of smoke obscuration are quantified at three locations, which are used to start the tunnel ventilation at 4 min. This can be related to a video-based smoke detection. The smoke model would be useful in tenability, egress and other life safety assessments. Future development of the model includes local lighting effects and experimental validation.     

Adaptive flame detection using randomness testing and robust features

This paper presents a novel approach to detect flame based on robust features and randomness testing. The flame color probability is estimated based on a Gaussian model learned in the YCbCr color space. The motion probability is then obtained by employing the background image updated dynamically with an approximate median method. The color and motion probabilities are integrated in order to obtain flame candidates, from which a feature vector comprised of seven features is extracted for each frame. The successive feature vectors are then applied to the Wald–Wolfowitz randomness test in order to obtain the prior flame probability. Finally, the convolution is defined in order to update the prior probability into a posterior probability for improving the system reliability, and an alarm level is determined based on the posterior probability. The presented method was successfully applied to real-environment intelligent surveillance systems and proved to be effective, robust, and adaptive, irrespective of the environment, weather conditions, or video quality.     

Automated Analysis of Mobile LiDAR Data for Component‐Level Damage Assessment of Building Structures during Large Coastal Storm Events

Rapid assessment of building damages due to natural disasters is a critical element in disaster management. Although airborne‐based remote sensing techniques have been successfully applied in many postdisaster scenarios, automated building component‐level damage assessment with terrestrial/mobile LiDAR data is still challenging to achieve due to lack of reliable segmentation methods for damaged buildings. In this research, a novel building segmentation and damage detection approach is proposed to realize automated component‐level damage assessment for major building envelop elements including wall, roof, balcony, column, and handrail. The proposed approach first conducts semantic segmentation of building point cloud data using a rule‐based approach. The detected building components are then evaluated to determine if the components are damaged. The authors applied this method on a mobile LiDAR data set collected after Hurricane Sandy. The results demonstrate that the approach is capable of achieving 96% and 86% parsing accuracy for wall façades and roof facets, and obtain 82% and 78% of detection accuracy for damaged walls and roof facets.     

Transformation of a 4D product and process model to generate motion of mobile cranes

Modeling and visualization of motion of mobile cranes enable project engineers to identify possible spatial conflicts related to cranes prior to actual operation on job sites and hence, could minimize hazardous conditions and delays associated with spatial conflicts. Current tools for visualizing equipment motion in three dimensions (3D) and across time have limitations due to their reliance on users in modeling a sequence of movement of each piece of equipment experientially and manually. This paper presents an approach for automatically generating motions of mobile cranes to support conflict detection. This approach builds on and extends existing approaches in product and process modeling and visualization of construction operations. It takes a product and process model that contains building design and schedule information of a specific project together with specifications of cranes and construction methods utilized by the project as input. The crane specifications and construction methods constitute project-independent information that describes how cranes should be operated during the execution of activities. The output of the approach is a transformed product and process model that incorporates a set of operations and motions of cranes, which can be used for identifying spatial conflicts associated with crane operations. Validation studies show that the developed approach can be used to model different types of mobile cranes and generate their motion during operations, which enables detection of spatial conflicts related to cranes.     

Fire Detection in Video Using LMS Based Active Learning

In this paper, a video based algorithm for fire and flame detection is developed. In addition to ordinary motion and color clues, flame flicker is distinguished from motion of flame colored moving objects using Markov models. Irregular nature of flame boundaries is detected by performing temporal wavelet analysis using Hidden Markov Models as well. Color variations in fire is detected by computing the spatial wavelet transform of moving fire-colored regions. Boundary of flames are represented in wavelet domain and irregular nature of the boundaries of fire regions is also used as an indication of the flame flicker. Decisions from sub-algorithms are linearly combined using an adaptive active fusion method. The main detection algorithm is composed of four sub-algorithms (i) detection of fire colored moving objects, (ii) temporal, and (iii) spatial wavelet analysis for flicker detection and (iv) contour analysis of fire colored region boundaries. Each algorithm yields a continuous decision value as a real number in the range [−1, 1] at every image frame of a video sequence. Decision values from sub-algorithms are fused using an adaptive algorithm in which weights are updated using the least mean square (LMS) method in the training (learning) stage.