Video Flame Research Articles

Video flame recognition based on LGATP texture feature and sparse representation

Video flame recognition based on LGATP texture feature and sparse representation

Flame suppression due to acoustic streaming by using time reversal

Using high amplitude time reversal focusing to extinguish small flames is a novel demonstration of transient acoustic streaming. Typically, others have placed a low frequency acoustic source near the flame in order to generate a high enough particle displacement to extinguish it. In this study, we move the sources far from the flame and demonstrate that the time reversal method can focus transient waves to the flame. The peak acoustic overpressure level needed to extinguish a candle flame in free space is 191 dB peak when using a frequency range of 300 to 15000 Hz. This momentary focus causes acoustic streaming at the flame location after the focusing, which extinguishes the flame. By tracking the flame in high-speed video, we show the displacement of the flame due to the passing acoustic wave and subsequently due to the acoustic streaming. We further show that acoustic streaming extinguishes the flame, not the acoustic particle displacement.

Effective flame detection and classification using deep feature of QuickDenseNet and Ensemble Score Voted SVM

Fire poses a significant threat to both lives and property, necessitating effective early detection measures. Despite challenges in identifying smoke and fire in their initial stages, we have devised a cost-efficient visual detection system. Early fire detection enhances its potential effectiveness. CCTV surveillance systems are now commonplace in developed countries, serving as tools for periodic monitoring of various locations. However, fluctuating ambient light conditions, camera angles, and seasonal variations can introduce data distortions, occlusions, and impact model accuracy. To address these issues, we’ve implemented a method combining deep learning networks and machine learning strategies for flame detection and direction classification. Our innovative QuickDenseNet extracts dense features from segmented flame video frames. We introduce the Ensemble Score Voted SVM (ESV-SVM), employing SVM as the primary learner and score voting as the auxiliary learner. Our approach is rigorously evaluated through simulations, measuring accuracy and various Key Performance Indices (KPIs), including Precision, F1-score, Recall, Correlation, Error, FPR, and Correlation Coefficients. Remarkably, our proposed method achieves an impressive precision rate of approximately 99.5%.

Online Combustion Status Recognition of Municipal Solid Waste Incineration Process Using DFC Based on Convolutional Multi-Layer Feature Fusion

The prevailing method for handling municipal solid waste (MSW) is incineration, a critical process that demands safe, stable, and eco-conscious operation. While grate-typed furnaces offer operational flexibility, they often generate pollution during unstable operating conditions. Moreover, fluctuations in the physical and chemical characteristics of MSW contribute to variable combustion statuses, accelerating internal furnace wear and ash accumulation. Tackling the challenges of pollution, wear, and efficiency in the MSW incineration (MSWI) process necessitates the automatic online recognition of combustion status. This article introduces a novel online recognition method using deep forest classification (DFC) based on convolutional multi-layer feature fusion. The method entails several key steps: initial collection and analysis of flame image modeling data and construction of an offline model utilizing LeNet-5 and DFC. Here, LeNet-5 trains to extract deep features from flame images, while an adaptive selection fusion method on multi-layer features selects the most effective fused deep features. Subsequently, these fused deep features feed into DFC, constructing an offline recognition model for identifying combustion status. Finally, embedding this recognition system into an existing MSWI process data monitoring system enables online flame video recognition. Experimental results show remarkable accuracies: 93.80% and 95.08% for left and right grate furnace offline samples, respectively. When implemented in an online flame video recognition platform, it aptly meets recognition demands.

A Real-Time Flame Detection Method Using Deformable Object Detection and Time Sequence Analysis.

Timely and accurate flame detection is a very important and practical technology for preventing the occurrence of fire accidents effectively. However, the current methods of flame detection are still faced with many challenges in video surveillance scenarios due to issues such as varying flame shapes, imbalanced samples, and interference from flame-like objects. In this work, a real-time flame detection method based on deformable object detection and time sequence analysis is proposed to address these issues. Firstly, based on the existing single-stage object detection network YOLOv5s, the network structure is improved by introducing deformable convolution to enhance the feature extraction ability for irregularly shaped flames. Secondly, the loss function is improved by using Focal Loss as the classification loss function to solve the problems of the imbalance of positive (flames) and negative (background) samples, as well as the imbalance of easy and hard samples, and by using EIOU Loss as the regression loss function to solve the problems of a slow convergence speed and inaccurate regression position in network training. Finally, a time sequence analysis strategy is adopted to comprehensively analyze the flame detection results of the current frame and historical frames in the surveillance video, alleviating false alarms caused by flame shape changes, flame occlusion, and flame-like interference. The experimental results indicate that the average precision (AP) and the F-Measure index of flame detection using the proposed method reach 93.0% and 89.6%, respectively, both of which are superior to the compared methods, and the detection speed is 24-26 FPS, meeting the real-time requirements of video flame detection.

Fully Synthetic Videos and the Random-Background-Pasting Method for Flame Segmentation

Video-based flame detection (VFD) aims to recognize fire events by using image features. Flame segmentation is an essential task in VFD, providing suspected regions for feature analysis and object recognition. However, the lack of positive flame samples makes it difficult to train deep-learning-based VFD models effectively. In this paper, we propose the assumption that we can train a segmentation model with virtual flame images and design experiments to prove it. We collected many virtual flame videos to extend existing flame datasets, which provide adequate flame samples for deep-learning-based VFD methods. We also apply a random-background-pasting method to distribute the flame images among different scenarios. The proposed method trains a flame segmentation model with zero real flame images. Moreover, we perform segmentation testing using real flame images, which the model has never used, to see if the model trained using ‘fake’ images can segment real objects. We trained four segmentation models based on FCN, U-Net, Deeplabv3, and Mask-RCNN using synthetic flame video frames and obtained the highest mPA of 0.783 and mIoU of 0.515. The experimental results on the FIRE-SMOKE-DATASET and the Fire-Detection-Image-Dataset demonstrate that the ‘fake’ flame samples generated by the proposed random-background-pasting method can obviously improve the performance of existing state-of-the-art flame segmentation methods using cross-dataset evaluation settings.

Video flame recognition based on α-GMM and weight kernel sparse representation

Aiming at the problems of incomplete extraction on candidate flame region, a video flame recognition method is proposed based on Gaussian mixture model with adaptive learning rate α (α-GMM) and weight kernel sparse representation. Firstly, a Gaussian mixture model (GMM) for foreground extraction is constructed. Its learning rate α is adjusted adaptively according to the complexity of background change, and candidate flame region is extracted completely by combining the color features in the HSI space. Secondly, dynamic and static features of the flame are extracted from the candidate region, and a feature dictionary is constructed. Finally, a weighted kernel sparse representation classification model based on Mahalanobis distance (MD) is established to implement flame recognition. The kernel function is adopted to solve the nonlinear distribution problem effectively. In order to strengthen the discrimination between classes and improve the flame recognition rate, MD is employed to measure the similarity information between data and construct the weight matrix. The experiment results show that the candidate flame region obtained by the proposed α-GMM is more complete, and the segmentation accuracy is higher. Compared with other classifiers, the accuracy of the proposed weight kernel sparse representation classifier is improved by 10.94% on average, and the false positive rate is reduced by 55.19% on average, which indicates that the proposed method has a higher recognition rate and strong robustness.

Combustion Condition Recognition of Coal-Fired Kiln Based on Chaotic Characteristics Analysis of Flame Video

Keeping combustion stable and detecting unstable states in time is crucial for coal-fired furnaces such as rotary kilns, boilers, and oxygen furnaces. Because of the interference and complex conditions in the industrial field, recognition of combustion conditions by vision analysis is difficult. In this article, we propose a robust nonlinear dynamic system analysis-based approach for combustion condition recognition by extracting chaotic characteristics from a flame video. We first discover chaotic characteristics in the intensity sequence extracted from a flame video of coal-fired kilns, and then we further find that the underlying chaos rules differ between combustion conditions. Based on this finding, we design a set of trajectory evolution features and morphology distribution features of chaotic attractors for combustion condition recognition. After reconstructing the chaotic attractors from the intensity sequence of a flame video by phase space reconstruction, the quantified features are extracted from the recurrence plot and morphology distribution and put into a decision tree to recognize the combustion condition. The experimental results on real-world data show that the proposed method can recognize the combustion condition in coal-fired kilns effectively and promptly. Compared with other methods, the recognition accuracy is improved more than 5%.

Multi-Scale Video Flame Detection for Early Fire Warning Based on Deep Learning

The widespread use of renewable energy resources requires more immediate and effective fire alarms as a preventive measure. The fire is usually weak in the initial stages, which is not conducive to detection and identification. This paper validates a solution to resolve that problem by a flame detection algorithm that is more sensitive to small flames. Based on Yolov3, the parallel convolution structure of Inception is used to obtain multi-size image information. In addition, the receptive field of the convolution kernel is increased with the dilated convolution so that each convolution output contains a range of information to avoid information omission of tiny flames. The model accuracy has improved by introducing a Feature Pyramid Network in the feature extraction stage that has enhanced the feature fusion capability of the model. At the same time, a flame detection database for early fire has been established, which contains more than 30 fire scenarios and is suitable for flame detection under various challenging scenes. Experiments validate the proposed method not only improves the performance of the original algorithm but are also advantageous in comparison with other state-of-the-art object detection networks, and its false positives rate reaches 1.2% in the test set.

Gabor-GLCM-Based Texture Feature Extraction Using Flame Image to Predict the O2 Content and NO x.

Flame image feature extraction is the basis for boiler combustion monitoring and control. The flame video images of recent research are mainly derived from experimental burners in the laboratory, and few pay attention to the flame images in industrial boilers. The actual industrial boiler flame images differ significantly from the laboratory flame images. Additionally, certain flame image features cannot be captured in the laboratory owing to the limitations of the camera installations. Therefore, a flame image texture feature extraction algorithm based on an industrial boiler is proposed in this paper. The texture features were enhanced using a Gabor filter for the RGB channels of the flame images, and then, the statistics of the texture features were scalarized by a gray-level co-occurrence matrix (GLCM). The data were filtered and downscaled by a data compressor consisting of Gaussian-weighted mean and principal component analysis (PCA) to obtain eight key variables. The extracted eight variables were verified to be effective in characterizing the O2 and NOx contents of flue gas using the mutual information method. The combustion process regression model was constructed using a gated recurrent unit (GRU) on the 8 h combustion data of the boiler, and the predicted mean absolute percentage error (MAPE) for O2 and NOx content in the test set reached 7.5 and 10.2%, respectively. Compared to the conventional methods of direct PCA on images and GLCM plus PCA on images, the MAPE for O2 content prediction was reduced by 12.3 and 7.3%, and the MAPE for NOx content prediction was reduced by 10.5 and 6.1%, respectively. The advantage of the new flame feature based on Gabor-GLCM is suitable for the subsequent analysis and control of an industrial combustion system.

Early detection of lean blowout in a combustor using symbolic analysis of colour images

Recently, lean combustion has been accepted as one of the preferable modes to run both small and large scale energy sectors as it reduces the level of NOx (oxides of nitrogen) emissions. However, ultra-lean combustion is more susceptible to lean blowout (LBO) which significantly reduces engine life and operating reliability. Therefore, early detection of LBO is important so that operator can get sufficient lead time to initiate precautionary actions. The current investigation proposes a novel technique based on the conversion of flame colour matrix into symbolic strings. The generation of symbolic strings is done based on the partitioning of the hue space obtained from transformation of RGB (Red-Green-Blue) to HSV (Hue-Saturation-Value) using uniform partitioning method. The partitions are determined at the reference state and kept fixed for other equivalence ratios (anomalous or non-reference states). Thus, the hue value at each pixel of the image is assigned a symbol depending on its belonging to a particular partition. Thereafter, an array of symbolic state probability vector at reference combustion state is determined based on the probability of pixel hue value being in a particular symbolic state or symbol (number of symbolic states is considered as 8). Also, for other anomalous or non-reference combustion states, symbolic state probability vectors are estimated using the partition formed at the reference state. The difference in symbolic state probability vectors between reference and anomalous combustion states gives the anomaly measure, which is used here to detect the flame blowout. The early increasing trend of anomaly towards the blowout for both premixed and partially premixed flames indicate that the measure can be helpful to detect the approach towards LBO. The estimation of anomaly is shown for using the flame images of different quantities. The random selection of five images from a flame video shows to be very effective in capturing the trend of anomaly which also ensures low computational effort. Thus, the tool based on symbolic colour image analysis (SCIA) may also be helpful in the mitigation of LBO.

Study on threshold selection method of continuous flame images of spray combustion in the low-pressure chamber

The study of combustion in the plateau environment has gradually aroused the interest of scholars. Due to the particularity of the plateau environment, these plateau combustion experiments are usually carried out in artificial low-pressure chambers. The image processing of continuous flames has recently been widely used in scientific research and detection applications for various thermal and power devices. In this paper, three methods for selecting the optimal threshold for flame binarization images were investigated by using a flame video of horizontal diesel spray combustion in a low-pressure chamber. The results show that the binarization image with the threshold value selected by the two-peak method came closer to the actual flame's morphology. The maximum class difference method greatly increases the binarization area at the bottom of the flame, while the minimum error method significantly reduce the binarization area of the flame, especially with regard to the flame length. Finally, the mechanism of influence of the threshold selection method was analyzed from the perspective of the flame luminance and buoyancy of the horizontal jet flame. This work provides a better method of threshold selection for scientific research into simulating plateau combustion flames in low-pressure chambers, which has important engineering value and significance.

Deep-Learning Technique Monitors Well-Testing Burner Efficiency

This article, written by JPT Technology Editor Chris Carpenter, contains highlights of paper IPTC 19714, “Deep Learning for Monitoring Well-Testing Burner Efficiency,” by Darine Mansour, SPE, Hakim Arabi, and Yingwei Yu, SPE, Schlumberger, et al., prepared for the 2020 International Petroleum Technology Conference, Dhahran, Saudi Arabia, 13-15 January. The paper has not been peer reviewed. Copyright 2020 International Petroleum Technology Conference. Reproduced by permission. During well-testing operations, incorrect burner combustion could have an adverse effect on the environment or personal safety. Combustion efficiency is assessed by personnel who observe the flame. This practice lacks consistency and poses challenges, including environmental and safety considerations and issues with data. In the complete paper, the authors propose a solution that uses a deep neural network that learns from flame videos to define the quality of the combustion. The results are promising and show that this solution is a good candidate for real-time burner-efficiency monitoring and automatic alarm triggering and optimization. Introduction Automation of the evaluation of combustion efficiency would promote consistent combustion efficiency and timely reaction to undesirable combustion. In field operations, cameras are positioned for a direct view of the burner flames. The authors use a data set of video images captured by the cameras and labeled by domain experts as images of acceptable or undesirable combustion. The goal of the study is to automate human monitoring of the combustion by processing these images and classifying them automatically as images of acceptable or undesirable combustion. A supervised learning approach is applied to the problem. The approach consists of extracting the flame features from each image and feeding the features to a classifier that categorizes the combustion. The classifier is trained using the extracted features and labeled combustion images as input. The traditional approach to feature extraction is manual feature engineering, involving the extraction of handcrafted features using image-processing algorithms and domain knowledge. The process is tedious and time-consuming. Much effort is required to extract all relevant features in various contextual and environmental conditions. The extracted features are problem-dependent, and the process may require revisiting algorithms to adapt to new data that hold new features or new representations of existing features. Approach To achieve classification, a convolutional neural network (CNN) is trained using a data set of images labeled by domain experts. Using the image data set captured by well-testing cameras, a CNN with an architecture similar to VGG-16 that learns to extract powerful features from natural images is fine-tuned.

Automatic Smoke Detection Based on SLIC-DBSCAN Enhanced Convolutional Neural Network

Video flame and smoke-based fire detection usually exhibit large variations in the feature of color, texture, shapes, etc., caused by the complex environment. It is difficult to develop a robust method to detect fire based on single or multiple fire features. Since convolutional neural network (CNN) has reported state-of-the-art performance in a wide range of fields. This study present a method based on SLIC-DBSCAN and convolutional neural network to recognize flame and smoke modes connected to fire stages. First, simple linear iterative clustering (SLIC) is acted as the pre-processing step to over segment images into super-pixels. Then the use of density based spatial clustering of application with noise (DBSCAN) gathered the similar super-pixels into several clusters, which in turn provide better smoke detection accuracy by using CNN. Comparison studies are performed to base on smoke image from publicly available data and self-collected data. The experimental results demonstrated the improved smoke detection capabilities by the present method.

基于改进GMM和多特征融合的视频火焰检测算法

基于改进GMM和多特征融合的视频火焰检测算法

Prognostics of Combustion Instabilities from Hi-speed Flame Video using A Deep Convolutional Selective Autoencoder

The thermo-acoustic instabilities arising in combustion processes cause significant deterioration and safety issues in various human-engineered systems such as land and air based gas turbine engines. The phenomenon is described as selfsustaining and having large amplitude pressure oscillations with varying spatial scales of periodic coherent vortex shedding. Early detection and close monitoring of combustion instability are the keys to extending the remaining useful life (RUL) of any gas turbine engine. However, such impending instability to a stable combustion is extremely difficult to detect only from pressure data due to its sudden (bifurcationtype) nature. Toolchains that are able to detect early instability occurrence have transformative impacts on the safety and performance of modern engines. This paper proposes an endto- end deep convolutional selective autoencoder approach to capture the rich information in hi-speed flame video for instability prognostics. In this context, an autoencoder is trained to selectively mask stable flame and allow unstable flame image frames. Performance comparison is done with a wellknown image processing tool, conditional random field that is trained to be selective as well. In this context, an informationtheoretic threshold value is derived. The proposed framework is validated on a set of real data collected from a laboratory scale combustor over varied operating conditions where it is shown to effectively detect subtle instability features as a combustion process makes transition from stable to unstable region.

Burning condition recognition of rotary kiln based on spatiotemporal features of flame video

In the coal-fire industry, recognition of burning condition is vital for combustion control and optimization as it can provide early warning of abnormal conditions in the combustion system. We propose an effective model based on spatiotemporal features extracted from flame video for burning condition recognition especially for unsteady state condition. This model extracts features from both the spatial and temporal dimensions based on multiple adjacent frames to capture the appearance and motion information of the flame. Two new three-dimensional (3D) descriptors are developed to characterize the dynamic characteristics of the flame video stream. A computationally simple and fast dynamic texture descriptor, 3DBLBP, is designed to extract flame dynamic texture and motion from three adjacent frames of a video block, and a dynamic structure descriptor, HOPC-TOP is developed by extracting 3D phase congruency information of video clip from three orthogonal planes to capture structure and motion characterizes of flame. The two descriptors are combined to extract spatiotemporal features from flame clip for burning condition recognition. Experiment results show the proposed framework can achieve a high recognition accuracy in the real-world data, thus verifying the effectiveness of our proposed framework for the recognition of burning conditions in a rotary kiln.

A Flame Detection Method Based on Fusion Feature and SVM for Substation Inspection

For the safety inspection of substation, this paper proposes a flame image detection method based on the combination of fusion feature and support vector machine (SVM). Firstly, the suspected flame area is detected by motion detection and flame color model. Then, four features of B-channel color variation coefficient, roughness, flicker frequency and flame area change rate are extracted. Finally, four features are sent to SVM to learn the flame classifier and used in video flame detection. The experimental results show that the proposed method can effectively identify the flame.

Research on video flame detection algorithm based on improved DS evidence theory

The detection and prevention of flame is of great value to protect people’s lives and property safety. At present, most flame detection methods use a single classifier and have achieved some results. However, a single classification algorithm has poor adaptability to fire detection in a variety of complex situations. Therefore, a multi-classifier fusion flame detection algorithm is proposed based on Dempster-Shafer (DS) evidence theory is proposed. In short, firstly four classifiers are used to classify the same flame feature, and the four classification results are fused to make preliminary decision. The four classifiers include support vector machine (SVM), K-nearest neighbor (KNN), decision tree (DT) and random forest (RF). Second, three complementary flame features are chosen, namely color, texture and shape changes. Finally, the preliminary decision results of the three features are fused to obtain the final classification result. It should be noted that when different classifiers have strong conflicts on the classification result of the same feature, the fusion rule of DS evidence theory will be invalid. To solve this problem, the DS evidence theory is improved. For the experiment, the public flame videos are collected to construct a data set including different complex scenes for algorithm verification, where the frame rate of the video is 15 or 24 frames/s and the resolution is 320 × 240. The experimental results show that there is a strong complementary among the results of different single classifier. The multi-classifier fusion algorithm can achieve better classification performance and robust performance than the single classifier by integrating the results of each classifier, and its average detection rate reaches 93.08%. In addition, for the changes of different environments, the proposed method has higher adaptability and stability than other state-of-art methods.

Indoor Video Flame Detection Based on Lightweight Convolutional Neural Network

At present, all CNN-based fire detection algorithms identify fire by means of a single frame image, all of which demonstrate low accuracy under strong interferences or complex backgrounds such as flickering light or backgrounds with high level of brightness. To increase the accuracy of fire detection, this paper presents a neural network model which combines lightweight CNN with SRU. In this algorithm, the scene content is extracted by CNN and the dynamic characteristics of the flames are extracted from sequential frames. In this paper, Resnet18+SRU (V1-type) and Mobilenets+SRU (V2-type) are proposed. Based on the characteristics of flames at a fixed position within a short period of time, a 3D convolutional layer is added between the Mobilenets and the SRU in the V2-type model, resulting in the V3-type model. Based on a cross validation set containing multiple types of interference in an indoor environment, experiments were conducted to compare the three models proposed in this paper with other models. The experiment results showed that the accuracy of the method proposed in this paper is above 96%, about 25% higher than the accuracy of CNN-based fire alarm via single-frame image, and that the V3-type models with 3D convolutional layer has the highest accuracy and best overall performance.