Search

Smoke image generation is an important method to solve sample categories imbalance in smoke detection with applications to forest smoke detection, smoky vehicles detection, etc. This article presents two Controllable Smoke Image Generation Neural Networks (CSGNet and CSGNet-v2). More specifically, CSGNet is proposed to generate various smoke images by integrating a smoke component control module (SCM) and a smoke image synthesis module (ISM) with a multi-scale strategy. By changing specified smoke components latent codes, we can generate smoke images with specified smoke components. By fine-tuning smoke components latent codes in SCM, we can fine-tune the smoke components in generated smoke images. To further improve CSGNet, CSGNet-v2 is proposed to make generated smoke images more realistic by adding a smoke image fine-tuning module (IFM). The experiment results show that our methods achieve by far the best results in generating smoke images with controllable smoke components.

During endoscopic surgery, smoke removal is important and meaningful for increasing the visual quality of endoscopic images. However, unlike natural image dehaze, it is practical impossible to build a large paired endoscopic image training dataset with/without smoke. Therefore, in this paper, we propose a new approach, called Desmoke-CycleGAN, which combined detection and removal of smoke together, to improve the CycleGAN model for endoscopic image smoke removal. The detector can provide information about smoke locations and densities, which helps the GAN model to be more stable and efficient for smoke removal. Although some imperfections still exist, the experimental results have demonstrated that this method outperforms other state-of-the-art smoke removal approaches with unpaired real endoscopic images.Clinical Relevance- This can help improve the visibility in endoscopic surgery and to get smoke-free endoscopic images with better quality.

The number of forest fires has increased significantly in recent years. Interventions to forest fires from air and land are very difficult and dangerous. In addition, since these interventions are time-consuming, forest fire can spread to a much wider area. In this study, an automatic fire extinguishing system was designed using SCADA for 5 forest regions up to 5 hectares. In the event of a fire in the system, the smoke sensor in the fire area becomes active and the IP camera image in this area is displayed on the screen. The pressurized water flushing tap, which automatically rotates 360 degrees, works in the area. If desired, automatic operation can be canceled and manually adjusted water flushing tap can be operated. The system gives a warning when the tank water drops below 1 ton. The aim of this study is to automatically intervene early in any fire that may occur in the forest and to control the fire before it spreads.

Until now, the problem of cigarette smoke in public places is still a problem among the community. To solve this problem, this research will design and implement a device capable of detecting the presence of cigarette smoke and its temperature using the MQ-135 smoke sensor and the DS18B20 temperature sensor. If cigarette smoke is detected, the wifi module receives data on the Arduino, then the LCDs the smoke, temperature, defuzzification, and speed values on the fan simultaneously and the wifi transfer module on Android / Hp is displayed on the LCD and the fan. functions automatically to inhale or absorb cigarette smoke in the room and cigarette smoke is filtered out through the air neutralizer so that the air that comes out of the room is clean.

This paper presents a method that aims to achieve physics-based smoke rendering in VR despite its hardware limitations. This problem is addressed by decoupling the smoke animation and rendering into independent stages. Offline smoke diffusion simulation was conducted in respect to fluid dynamics. Real-Time rendering in VR was achieved by the combination of the simulation's accumulated data and light scattering theory. Results demonstrate that a satisfying framerate is accomplished making this approach suitable for developing immersive and interactive applications such as training simulations.

According to research worldwide, dense smoke usually spreads faster than flames when a fire occurs. This situation often causes firefighters to feel “spatially confused”, making them run out of oxygen, or escapers cannot see the escape route and inhale harmful smoke. This situation always causes a regretful incident. Therefore, this research has developed a negative-ion smoke-removal helmet. We respectively installed direct current and alternating current negative ion emitter modules on the front edge of the helmet, then put the helmet in a transparent smokebox to observe its PM2.5 removal ability in unit space. During the experiment, we set up a camera outside the transparent smokebox to record visual changes in the external environment and a small webcam inside the helmet to record the changes in visibility inside the helmet simultaneously. Finally, when the PM2.5 value no longer drops, the smoke removal process ends. Then, through the changes of the RGB value of the internal image record, the helmet visibility change ratio between the start and the end of the smoke removal process is calculated. At the same time, to explore the best mode of the negative ion emitter array installed in a building staircase, we installed a negative ion emitter with a small fan on the wall of a building's staircase to record the corresponding data between the factors of emitter power, emitting height, wind speed, negative ion concentration, negative ion scattering radius. Through this research, we hope to provide better protection for escapers and firefighters in a fire scene.

Smoke recognition has been actively studied in the computer vision domain. Due to large variance of smoke color, texture and shapes, smoke recognition is a challenging task. Traditional smoke recognition methods are based on handcrafted features. In the past few years, some methods which are based on convolutional neural networks have been proposed that achieved great improvement of smoke recognition. However, previous methods cannot capture the internal structure of smoke well. Manifolds can represent the internal characteristics of smoke data in lower dimensions, reduce the redundancy of data representation, and obtain more discriminative capabilities. In this paper, an end-to-end deep convolutional neural network which based on manifold structure was proposed to capture more discriminative features for smoke recognition. Experimental results show that the proposed method achieves promising results on the public smoke recognition dataset. The proposed method can obtain high detection rate, high accuracy rate and low false alarm rate in the same time.

Sheathing materials of the power cables play an important role in the protection of the cable against mechanical damage and fire accidents. The sheathing materials are selected based on the application and the installation environment of the cables. Generally used sheathing materials for power cables are Poly Vinyl Chloride (PVC), Flame Retardant Poly Vinyl Chloride (FR PVC), Flame Retardant Low Smoke Poly Vinyl Chloride (FRLS PVC) and Low Smoke & Zero Halogen (LSZH) materials. PVC possesses high mechanical strength, and electrical strength. However, the smoke and toxic gas release during fire is more for PVC. To improve the flame retardancy and smoke properties, additives and fillers are blended to PVC to make it as FR PVC and FRLS PVC. The optimal selection of additives is done in such a way that the materials retain its basic mechanical and electrical properties along with improved fire characteristics. LSZH materials emit low smoke and no halogens. It is having high amount of mineral fillers to enhance the fire and smoke properties. Lubricants, stabilizers, plasticizers, antioxidants, and colorants are also added to retain the flexibility of the cables, avoid the deterioration of the materials and to achieve the desired colors. In the present study, the various sheathing materials of new power cables are analyzed for its mechanical and fire properties with ageing. The sheathing materials are aged at its field operating temperatures for duration of up to 672 hours and the properties are measured periodically. Ageing of the special sheathing materials have not resulted into significant changes in the fire properties except the loss of chlorine in PVC & FRLS PVC. However, the ageing of the sheathing materials has resulted into loss of certain mechanical properties. Hence it is suggested to evaluate the properties of sheathing materials with accelerated ageing at the compounding stage itself for its optimal compounding blend and to ensure the integrity of the compounds throughout its lifetime.

Several Internet of things (IoT) based smoke or fire detection designs have been made in previous research and a study has shown that the IoT system can implement fire sprinklers using a water pump controlled by microcontroller. However, control systems with proven methods such as Fuzzy Inference System (FIS) need to be implemented in the system and evaluated to improve the control system. This research focuses on designing and evaluating a smart fire sprinkler device prototype that can detect and extinguish fires based on the FIS classification and send the information to an android application using IoT. The FIS classification proposed in this study uses Tsukamoto inference based on several inputs taken from three sensors, which are a fire sensor, a temperature sensor, and a smoke sensor. Four fire conditions represent four different levels of sprinkler output: smoke, small fire using burnt straws, medium fire with a candle, and large fire using burnt paper balls. Through testing, the success rate of the system in classifying fire and extinguishing the fire has a macro average F1 score of 90%.

2020 was the most active wildfire season in recent history. This study leverages solar models to quantify the solar impacts of wildfire smoke in western North America 2001 – November 2020. We observe a sharp increase in the number of days impacted by aerosol events. Record deviations in clear sky DNI are found at the Hanford, CA, Boulder, CO, and Desert Rock, NV study locations. Total sunlight (GHI) for September was diminished by up to 20% in some locations; California’s Central Valley and parts of the Columbia River Basin were hardest hit by the smoke. At the Hanford study location, the Aug.- Oct. 2020 deviations in modeled energy output totaled -5.9% of the historical annual average (2001-2019). The analysis demonstrates that wildfires are an important risk to production for solar projects in western North America.

As an important structure connecting rooms and atrium in large shopping malls, corridors have a huge impact on smoke movement but are rarely mentioned in previous studies. The purpose of this study is to investigate the impact of changes in the corridor’s physical boundary (corridor width and keep stand height) on the development of fire within a typical room-corridor-atrium model. A series of cases were carried out numerically by using Fire Dynamic Simulator (FDS).The result shows that: the keep stand height has a huge effect on the smoke migration path, asymmetric distribution of temperature, and the mass flow rate. With a certain heat release rate, a critical width was found to determine the maximum value of the temperature and spill plume of the atrium. Based on the above analyses, some smoke control measures were put forward and discussed. It indicates that the exhaust vent position and the keep stand height should better be given position 3 and non-dimensional height with 0.375 respectively. Besides, the keep stand was also introduced into Poreh’s spill plume model to verify the relationship between mass flow rates and keep stand height, the modified model shows a good linear relationship.

Video smoke detection is a promising method for early fire prevention. However, it is still a challenging task for application of video smoke detection in real-world detection systems, as the limitations of smoke image samples for training and lack of efficient detection algorithm. This paper proposes a method based on two state-of-the-art fast detectors, a single-shot multi-box detector, and a multi-scale deep convolutional neural network, for smoke detection using synthetic smoke image samples. The virtual data can automatically offer rich samples with ground truth annotations. However, the learning of smoke representation in the detectors will be restricted by the appearance gap between real and synthetic smoke samples, which will cause a significant performance drop. To train a strong detector with synthetic smoke samples, we incorporate the domain adaptation into the fast detectors. A series of branches with the same structure as the detection branches are integrated into the fast detectors for domain adaptation. We design an adversarial training strategy to optimize the model of the adapted detectors, to learn a domain-invariant representation for smoke detection. The domain discrimination, domain confusion, and detection are combined in the iterative training procedure. The performance of the proposed approach surpasses the original baseline in our experiments.

Detecting forest fire smoke during the initial stages is vital for preventing forest fire events. Recent studies have shown that exploring spatial and temporal features of the image sequence is important for this task. Nevertheless, since the long distance wildfire smoke usually move slowly and lacks salient features, accurate smoke detection is still a challenging task. In this paper, we propose a novel Attention Enhanced Bidirectional Long Short-Term Memory Network (ABi-LSTM) for video based forest fire smoke recognition. The proposed ABi-LSTM consists of the spatial features extraction network, the Bidirectional Long Short-Term Memory Network(LSTM), and the temporal attention subnetwork, which can not only capture discriminative spatiotemporal features from image patch sequences but also pay different levels of attention to different patches. Experiments show that out ABi-LSTM is capable of achieving best accuracy and less false alarms on different types of scenarios. The ABi-LSTM model achieve a highly accuracy of 97.8%, and there is 4.4% improvement over the image-based deep learning model.

Detecting smoke from visual sensors is crucial for fire warnings. However, visually detecting smoke is still challenging because the chrominance, shape, transparency, texture, motion, and so on, of smoke vary over a wide range. A deep neural network (DNN) could detect smoke with higher accuracy, but the models are too large to run on limited resource platforms. To make the models smaller and improve the accuracy, three novel strategies are proposed in this paper: 1) feed the networks with blocks rather than the original image; 2) feed the network with condensed data rather than videos to extract dynamic characteristics; and 3) the framework has two stages, the first stage focuses on every block to find thin smoke, and the second stage focuses on the ascending and expanding motions in videos. It is not an end-to-end network, and every subnetwork works on a small matrix, even the one at the second stage, which operates on the whole image. Therefore, the network based on these strategies, the slight smoke perceptual network (SSPN), is small enough to be integrated on a raspberry pi. The experiments demonstrate that SSPN outperforms existing traditional methods and the methods based on DNN and achieves the highest accuracy and the highest sensitivity.

In recent years, due to the impact of environmental protection, reducing harmful gas emissions and other policies, lithium battery, hydrogen oxygen fuel cell and other technologies are being applied on ships. With the continuous expansion of the application level of new energy, its security issues are more and more concerned by the society. In this system, sensor fusion algorithm, FPGA control core and can bus architecture are introduced to collect the gas concentration, voltage, current, stack temperature and other information inside the fuel cell system, as well as humidity, temperature, smoke and other information in the battery compartment environment. After preliminary processing by the slave processor, the master processor judges whether the fuel cell has fire hazard Hidden danger. The system can effectively avoid thermal runaway of fuel cell, detect thermal runaway early in a low-cost and recyclable way, and actively respond to reduce the loss, so as to ensure the safety of life and property of passengers on board.

Aiming at the problem of many background interference factors in the traditional forest fire smoke monitoring video based on digital images, a detection method that can eliminate interference factors such as clouds and skylines and its specific implementation methods are proposed. The article first obtains the early forest fire smoke source detection video from close-range aerial photography of multi-rotor aircraft, and then uses the principle of multi-modal mixture Gaussian background model to establish a background model to extract the background of the early forest fire smoke video. The features of the image are high-frequency fusion, the fusion image is compared with the high-frequency energy value of the corresponding background image, if the high-frequency energy of the local area is reduced, it is determined as a suspected smoke area, and the largest connected area of suspected smoke is marked. Through the simulation experiment of the above method, the experimental results show that the method can accurately detect the smoke in the video frame image, the false detection rate is low, and the influence of related interference factors on the detection accuracy of the early forest fire smoke video image can be eliminated.

Recent advances in unmanned aerial vehicles and camera technology have proven useful for the detection of smoke that emerges above the trees during a forest fire. Automatic detection of smoke in videos is of great interest to Fire department. To date, in most parts of the world, the fire is not detected in its early stage and generally it turns catastrophic. This paper introduces a novel technique that integrates spatial and temporal features in a deep learning framework using semi-supervised spatio-temporal video object segmentation and dense optical flow. However, detecting this smoke in the presence of haze and without the labeled data is difficult. Considering the visibility of haze in the sky, a dark channel pre-processing method is used that reduces the amount of haze in video frames and consequently improves the detection results. Online training is performed on a video at the time of testing that reduces the need for ground-truth data. Tests using the publicly available video datasets show that the proposed algorithms outperform previous work and they are robust across different wildfire-threatened locations.

Early fire detection and alarm are significantly important to reduce the losses caused by fire. Conventional methods in fire detection using smoke and heat detectors have disadvantages in accuracy, latency as well as the detection area. In this paper, we propose and implement a real-time fire detection solution for large area surveillance using the unmanned aerial vehicle with an integrated visual detection and alarm system. The system includes a low-cost camera, a light weight companion computer, a flight controller as well as localization and telemetry modules. To achieve real-time detection, Single Shot MultiBox Detector (SSD) algorithm is implemented as the heart of the system. We used MobileNets base model, which more efficient for mobile and embedded vision applications, instead of conventional VGG-16/ResNet model to achieve the mean average precision of 92.7% with the detection speed of 26 FPS.

According to the 2019 National Fire Statistics Analysis of the Fire Department of the Ministry of the Interior, “dense smoke obstruction” is the main reason for fire death after deducting the factors of “inadequate judgment and physical strength” such as disability and deep sleep. As toxic smoke spreads faster than flames, it often causes rescuers to lose their sense of direction until the oxygen runs out and prevent people from viewing an evacuation route hence inhale harmful smoke, which leads to regrettable incidents. Therefore, this research explores the smoke removal efficiency of a high-power Negative Ion Emitter Array (NIEA) in space. We created a 1:1 smoke chamber with high-concentration smoke to simulate the actual situation of the fire scene and installed high-power NIEA on the vertical or top surfaces of the smoke chamber. During the experiments, we recorded the changes in the concentration of negative ions, PM2.5, and the visibility to calculate the smoke removal efficiency of NIEA under different configuration modes. The expected goal of this research is that in the future, the NIEA could be pre-installed in the structures of buildings or temporarily mounted by rescuers along the escape routes to help people obtain cleaner air in a smoke-filled environment and to survive.

The spread of a fire is an event that can cause a serious harm to people and their belongings. In this study, we propose an artificial neural network as a method to detect fire and smoke. This study aims to investigate the effectiveness of our three sensors that are the inputs of our system, Smoke Density, Temperature and Carbon Monoxide sensors. These sensors with the help of artificial neural network will help estimate the appearance of fire, smoke or none of the them. In the Artificial Neural Network, the forward pass, back-propagation and the further forward pass will be applied as methods to update weights and reach the minimal error by iterating through the epochs. This research has reached an accuracy of 98.3%.

Most existing smoke detection methods have low detection accuracy in complicated scenes. We propose a YOLO-SMOKE model which shows strong robustness and high accuracy. We use mosaic data augmentation to avoid model overfitting which is caused by the simple background in the public datasets. We refine the residual module by using the efficient channel attention module and add dropblock after every convolutional layer. Then we modify the loss function. The YOLO-SMOKE model outperforms the original model by 4.91% mAP and it achieves 98.75% accuracy in the public smoke video test. The experimental results show that our method has better accurate and robustness.

In order to solve the problems of transformer fire accidents such as the traditional fire control system has simple function, single fire characteristic parameters and is prone to false alarm, underreporting and so on, a synthetic diagnosis method using multi-neural network and evidence theory for transformer accidents is presented. The monitoring and early warning of transformer fire is realized by collecting temperature, video (flame and smoke), CO, combustion sound and other information in real time, and then combining with network technology. The results of example analysis show that the accuracy of this method for transformer fire identification is 73%.

In the last decade, information technology has gained a rapid development in the education industry. Particularly on the university campuses, information infrastructure is becoming more common and classrooms are becoming more modern. The increase of information infrastructure, however, while creating a comfortable teaching environment for students, it also brings more security risks. Thus, the safety of the classroom has also attracted more attention. Most classroom fire monitoring systems currently used are part of the traditional wired fire monitoring system. It is susceptible to a variety of objective factors, such as aging circuits, and some problems such as failure of the fire detection system. To address these challenges, this paper presents a CC2530 based classroom wireless fire monitoring system, and designs a fire monitoring system suitable for classrooms. The hardware part monitors the temperature and smoke concentration in the classroom through sensors to determine whether a fire occurs. The software part uses the ZigBee protocol stack Z-Stuck to send the monitored data to the monitoring center, enabling staff to monitor the classroom environment in real time. Experimental results show that the combined software and hardware system are well adapted to a modern classroom with many electrical devices and dense personnel. At the same time, the system possesses the characteristic of real-time data collection, high positioning accuracy and large transmission rate, which can meet the standards and requirements in practical applications.

Fire simulations are used to evaluate the spread of fires and check the safety of environments. To make the simulation realistic, the fire simulation must consider numerous physical and chemical reactions with complicate and time-consuming computations. It limits the fire simulations being used for real-time prediction when there is a fire. This paper presents a real-time fire prediction approach based on fire simulations. Our approach uses fire simulation to collect training data for building a smoke spread model of fire. The model is used to predict the spread of fire real-time which can help online evacuation planning more accurate and realistic, such as help people avoid dangerous area.

The fire safety of offshore booster station has an extremely important impact on the reliable operation of the whole offshore wind farm. The article establishes the corresponding fire protection scheme around the offshore wind turbine group, offshore booster station and the on-land central control center of the offshore wind farm. Among them, each offshore wind turbine, offshore booster station and onshore centralized control center are equipped with independent fire alarm systems. The inspiratory smoke detector and point type composite smoke and temperature detector are mainly used to confirm the secondary alarm. At the same time, the fire extinguishing systems of offshore wind turbines, offshore booster stations and on-land control centers are: high-pressure CO 2 fire extinguishing system, fine water mist fire extinguishing system and water spray fire extinguishing system. Finally, combined with an engineering case, the fire protection scheme of an offshore wind farm is designed.

Fire is one of the disasters against the Safety of human life and property. Generally, early smoke features are more obvious than fire in the surveillance environment. However, due to the variability of smoke characteristics (e.g., color, shape) and the interference of smoke-like objects (e.g., clouds, rivulet, and fog), the main challenge of smoke detection is false alarms in real-world. To tackle this problem, an integrated method is proposed which combines HSV color space, background subtraction with Faster R-CNN. This method can enhance smoke feature, meanwhile, it can reduce disturbance from smoke-like objects. Furthermore, a dataset is created which are collected from surveillance cameras that are installed in the wild. Our experiments show that the integrated method is more accurate and robust than previous work.

With the rapid development of social economy, the frequency of fires is increasing day by day, and indoor fires in buildings are particularly harmful to humans. Aiming at the current situation, an indoor fire early warning system based on video surveillance and smoke sensing is proposed. Using video images to identify and monitor flames, and at the same time use smoke sensors to detect and prevent fires, and issue corresponding alarms to indoor fire conditions, which can increase the accuracy of fire identification and reduce waste of resources. It effectively reduces the false alarm rate and the false alarm rate, and improves the reliability of the entire fire warning system.

Identification of air pollution is a significant task for environmental control, manage, and policy decision. In traditional approach, chemical composition analysis is very costly to be applied frequently and largely, especially in developing countries. This paper proposes the use of CALIPSO satellite image to analyze the aerosol sources, highly linking with particulate matter sources, in Hanoi in the periods from 2016 to 2019. Other datasets including Hanoi land-cover map and the monthly average wind direction from MERRA-2 reanalysis were utilized to explain the spatial distribution of aerosol sources. The result shows that polluted continental/smoke accounted for the largest proportion with 40%, followed by polluted dust, smoke, dust and clean continental with a percentage of 35%, 14%, 6% and 5%, respectively. The monthly variation of the aerosol type shown a high frequency of elevated smoke in March, April and October meanwhile polluted continental/smoke was a peak in the dry season (November to March) and lower in the rainy season (May to September). The aerosol types were observed mostly at high attitude including polluted dust, polluted continental and elevated smoke could be related to long-range transport from other places to Hanoi. This study highlights the potentials of using CALIPSO products for identification of air pollution sources in Vietnam.

The objective of this study is to build smoke and pollution free green sustainable resort city for the active seniors and nature lovers and to examine the resultant effects on active life, good health, and longevity. The study employs econometric theoretical model to evaluate the effects of building smoke and pollution free green and sustainable resort city on longevity. The theoretical model shows that green space, reduction in pollution, smokes and environmental damages improve longevity. In addition, plant, trees, vegetation, natural and quality food, and nutrition grown in healthy, organic, and natural environment also contribute positively to good health and longevity. The study finds that physical exercise, walking, running, jogging, biking, and hiking also contribute to good health, active life, and longevity. Preventing measures in health, environmental damages and food chain contamination also contribute positively for improving longevity. This is perhaps one of the latest study that emphasizes on environmental quality, natural habitat for both humans and other species, living side by side and thrive in harmony, and in natural balance, in pollution and smoke free environment and enjoy the greatest pay off in the form happiness, active life style and longevity.

Through the analysis of smoke particle size characteristics and the principle of photoelectric scattering, this paper studies the photoelectric scattering smoke detection technology based on the narrow band Internet of Things, and designs the data acquisition and communication terminal of the perception layer. The author proposes a technical method of using dual-band anti-interference photoelectric scattering to sense fire, and achieves high sensitivity and low false alarm rate for fire smoke detection. In this paper, a low-power optimization strategy for the device's periodic sleep mode is also proposed to reduce system power consumption.

Fire can cause serious damage to the natural environment and social economy, if it is not intervened early. Therefore, an effective fire detection method is significant and helpful. In this paper, a fire detection method based on deep learning is proposed to detect fire and smoke. Firstly, the residual network structure is applied to extract depth feature of the image. The problem of shallow features easily disappearing is solved with improved ResNet-50. A network composed of multiple BiFPN modules is established for multi-scale feature fusion and enhancement. Intersection over Union and cross entropy are applied to predict the scope and category of boundary boxes. Finally, the prediction results are obtained by comparing the confidence of the bounding box. Experimental results show that this method performs better in running time and accuracy than the existing detection networks. The feasibility of this method is verified in the field of fire detection.

In recent years, the photovoltaic industry has developed vigorously where fire occurs frequently at photovoltaic power plants. This paper proposes an image defogging method based on parallel convolutional neural networks and a combined detection method for smoke and flame based on improved Faster R-CNN. Firstly, fire datasets used in the photovoltaic power plant scene are collected and produced, including two categories of smoke and flame. Then, one deep learning method is used to suppress the haze interference in the photovoltaic power plant scene. Finally, we use ResNet instead of VGG16 network, and integrate feature pyramid network(FPN) into regional candidate networks(RPN) to improve the original Faster R-CNN structure. A large number of experimental results show that our improved model has better effects on joint detection of smoke and flame.

Aiming at the problems of traditional smoke detectors such as wiring difficulties, high line construction cost, weak anti-interference ability, and short transmission distance, this paper presents a design scheme of smoke detection and alarm system based on LoRa. The terminal is designed with the main control chip and LoRa module as the core, and the MCU is the control core to design the gateway, combined with the design of the communication protocol, to realize the functions of periodic data reporting, data abnormal alarm, abnormal state alarm and so on. Based on the LoRa network, this paper designs the overall architecture of the system. In addition, the article designs the system software, elaborates the system workflow, and proposes a low power consumption strategy for the terminal, which effectively reduces the system power consumption.

Industrial gas leakage in chemical plants presents a high risk to health and safety, also makes a huge waste of production materials. In this paper, we propose an automatic gas segmentation method for optical gas imaging (OGI) videos, which can not only indicate the existence of leakage in frames but also show the precise locations and shapes of gas plumes. Specifically, we propose a novel video segmentation network that stacks 2D spatial convolutions, 1D temporal convolutions and 3D spatial-temporal convolutions together within an encoder-decoder structure, termed 2.5D-Unet. Stacking mixed convolutions increases the representation ability of network for leakage’s appearance and motion. More importantly, stacking mixed convolutions facilitates pre-training of model using still images of smoke, which is especially useful when lacking of labeled videos of leaked gas. Experimental results show, for OGI video-based gas leakage segmentation, our method outperforms existing video segmentation methods.