672 results about "Fire detection" patented technology

The invention relates to the technical field of fire safety monitoring, in particular to a firework identification system and method based on multi-modal fusion perception, and the core of the scheme is a visible light, multispectral and temperature three-modal framework: feature extraction optimization of each modal, improved YOLOv8s for visible light branches, dynamic background modeling and flame color screening, and multi-modal fusion perception. False positive is rejected by a multispectral branch depending on a waveband ratio and an index, an error compensation algorithm is introduced into a thermopile branch, and finally, a final smoke and fire area and the confidence coefficient thereof are determined by associating a three-mode area through collaborative decision. According to the scheme, the complex environment adaptability and the recognition reliability can be improved, the false alarm risk is reduced, the extremely-early smoke and fire detection capability is enhanced, good real-time performance and deployment flexibility are achieved, and the method is suitable for various types of fire safety monitoring scenes.

The invention discloses an intelligent fire detection method based on artificial intelligence video analysis, and particularly relates to the technical field of computer vision. The method comprises the following steps: acquiring a video monitoring stream, extracting spatial color features and morphological features of an image, constructing a multi-scale time sequence feature graph group, performing joint modeling on color disturbance and structural disturbance by using a dual-channel attention neural network, and outputting a suspected fire area probability graph; candidate fire source areas are screened in combination with boundary disturbance consistency analysis and red channel high-frequency fluctuation detection; the texture stability, the disturbance directivity and the historical smoke evolution characteristics are integrated through a multi-modal fusion judgment model, the fire state is finally judged, an alarm result is output, meanwhile, the system can mark the fire position and the development trend path, and visual tracking is achieved. The method has the advantages of high robustness, high sensitivity and low false alarm rate, and is suitable for intelligent early warning of early fire in a complex environment.

The invention discloses a multi-modal lightweight forest fire detection method suitable for a vertical take-off and landing fixed-wing unmanned aerial vehicle platform, and the method comprises the steps: carrying out the feature alignment and weight fusion of a visible light image and an infrared image, which are synchronously collected by an unmanned aerial vehicle, through a multi-modal early fusion method, and obtaining a fusion feature; carrying out lightweight down-sampling and multi-scale feature extraction on the fusion features to obtain a multi-scale feature set; performing cross-scale feature fusion on the multi-scale feature set by adopting a bidirectional feature pyramid network, and extracting dynamic change information to obtain time sequence features; performing multi-scale context enhancement on the time sequence features by using a spatial pyramid method based on cavity convolution, and performing global context and local texture modeling by using a lightweight Transform module to obtain high-level semantic features; and based on the high-level semantic features, a target detection frame, fire level classification, prediction uncertainty estimation and a pixel-level fire probability graph are generated in parallel through a multi-task output header.

The invention relates to the technical field of environment early warning, and discloses a holographic sensing multi-mode environment early warning control method and system, and the method comprises the steps: obtaining an original multi-mode signal, and constructing a multi-mode fire signal matrix; performing matrix decomposition operation of combustion dynamics constraint and flue gas diffusion constraint on the multi-modal fire signal matrix to obtain a fire basic characteristic matrix and a spatio-temporal evolution coefficient matrix; performing feature fusion on the fire basic feature matrix and the spatio-temporal evolution coefficient matrix to obtain a comprehensive fire feature vector; according to the method, multi-dimensional fire feature information such as temperature, smoke, gas, acoustics and infrared can be captured at the same time, the comprehensiveness and reliability of fire detection are improved, and the method is suitable for popularization and application. And the three-dimensional space coordinates of the fire source can be accurately determined.

The invention discloses a fire detection method and device based on multi-modal perception and D-S evidence theory fusion. The method comprises the following steps: collecting multi-modal perception big data at least comprising video image data and temperature sensing data in a monitoring area; performing fire visual feature analysis on the video image data to generate first basic probability distribution; performing fire temperature characteristic analysis on the temperature sensing data to generate second basic probability distribution; taking the first basic probability distribution and the second basic probability distribution as two independent evidence sources, and fusing by adopting a D-S evidence theory to obtain a fused third basic probability distribution; converting the third basic probability distribution into a fire occurrence probability for decision making; and when the fire occurrence probability exceeds a preset alarm threshold, determining that a fire occurs and triggering an alarm. According to the fire detection method, multi-modal sensing big data are synchronously collected and fused, multi-source uncertain information is processed and decided under the framework of the D-S evidence theory, and the early stage of fire detection is remarkably improved.

The invention belongs to the technical field of computer vision and fire detection, and particularly relates to a lightweight fire detection method based on bimodal image fusion, and the method comprises the steps: collecting an RGB visible light image and a Thermal thermal infrared image in the same scene, and forming a group of samples; inputting the sample into a feature extraction module to obtain an RGB exclusive feature map and a Thermal exclusive feature map; the feature extraction module comprises a backbone network, an RGB feature branch and a Thermal feature branch; inputting the RGB exclusive feature map and the Thermal exclusive feature map into a feature fusion enhancement module to obtain a fused feature map; the feature fusion enhancement module comprises an MSR-CSSA module; inputting the fused feature map into an adaptive detection output module to obtain a prediction result; the self-adaptive detection output module comprises a three-branch detection head and a dynamic weight generation module; according to the invention, the detection precision and the real-time performance can be effectively improved.

The invention discloses an HRR-based acoustic fire hazard type identification method, device and equipment and a medium, and relates to the technical field of fire hazard detection, and the method comprises the steps: obtaining an original acoustic signal through transmitting a high-frequency modulation ultrasonic wave and receiving a flame region reflection echo, obtaining channel impulse response data through channel estimation, and carrying out the channel impulse response data; flame features are extracted, a heat release rate curve is calculated and generated, finally fire type identification and fire behavior stage determination are completed based on the curve, dense smoke, shielding and illumination interference can be resisted, accurate quantification of the heat release rate is realized, fire type identification and fire behavior development stage determination can be finely carried out, and the accuracy of fire behavior determination is improved. And a stable and reliable technical support is provided for fire monitoring and emergency decision making.

The invention discloses a safety and disinfection integrated system based on AI analysis linkage control electromagnetic valves, and relates to the technical field of intelligent fire fighting, and the safety and disinfection integrated system comprises a flame detection camera, a network input and output module, a safety and disinfection integrated machine, an execution module and an acousto-optic alarm bell. According to the security and disinfection integrated system for linkage control of the electromagnetic valve based on AI analysis, the advanced AI visual recognition technology and an electromagnetic valve linkage control mechanism are deeply fused, the efficiency and the precision degree of fire safety management are greatly improved, flame characteristics are precisely recognized by means of a deep learning model, and the safety and disinfection effect is improved. The limitation that traditional fire behavior detection is easily influenced by environmental factors is broken through, the accuracy and reliability of fire behavior recognition are greatly enhanced, the system supports multi-modal data fusion, the confidence coefficient of fire behavior judgment is further improved, adjacent electromagnetic valves can be automatically started according to the position of a fire behavior in the linkage control link, an effective fire extinguishing isolation belt is constructed, and the fire extinguishing efficiency is improved. And fire spreading is effectively restrained.

The invention relates to an emergency evacuation path generation method, system and device under multi-target conflict and a medium. The method comprises the following steps: acquiring video monitoring data and environment sensing data of a target building, and performing fire disaster detection according to the video monitoring data to obtain a fire detection result; in response to the fire detection result that the fire disaster exists, performing fire evolution prediction according to the environment sensing data based on the environment topological graph of the target building to obtain fire evolution parameters; performing risk assessment and latest passable time marking on each space unit node in the environment topological graph according to the fire evolution parameters to obtain a risk graph layer; and in response to the obtained positions of the to-be-evacuated persons, generating an exclusive emergency evacuation path of each to-be-evacuated person in combination with the environment topological graph and the risk map layer. According to the method, fire detection and fire spreading deduction are carried out through continuous video image frames, time sequence judgment is carried out on traffic effectiveness, and path design is prevented from lagging behind risk change.

The invention discloses an intelligent detection method and system for an abnormal edge of a charging station. The method comprises the following steps: S1, synchronously acquiring abnormal related RGB image data, infrared image data, ultraviolet image data and electrical and environmental data in a charging station scene; s2, visual target detection: S1, inputting the acquired RGB image into a lightweight target detection model, positioning a key target, outputting a bounding box, and providing an accurate region of interest (ROI) for subsequent anomaly recognition; s3, visual modal anomaly detection: performing multi-branch visual modal anomaly detection on the basis of the ROI obtained in the S2 and the target category of the ROI; s4, electrical-environment modal anomaly detection: detecting electrical anomaly and environment anomaly through a lightweight anomaly detection model according to the electrical and environment sequence data collected in the S1; and S5, multi-modal fusion abnormity identification alarm: based on the abnormity detection results in the S3 and S4, adopting a multi-modal fusion method to identify abnormity. According to the invention, a new two-stage anomaly detection framework fully utilizing multiple modes is provided, and comprehensive abnormal conditions such as charging pile damage, charging gun damage, parking space occupation of a fuel truck, abnormal user behavior and fire detection can be detected; in order to solve the problem of difficulty in edge deployment, the model is lightweight as much as possible while the performance is ensured.

The invention provides a cable-type temperature-sensing fire detection method and system with an intelligent positioning function, and relates to the technical field of artificial intelligence, and the method comprises the steps: laying an optical fiber temperature-sensing cable along a preset path, obtaining the real-time three-dimensional coordinates of a temperature measurement point through employing a reflective mark point and a multi-view three-dimensional reconstruction technology, enabling the temperature data to be correlated with the coordinates to form a temperature field data set, and obtaining a temperature field data set; and generating three-dimensional temperature field grid data by adopting a weight optimization method based on distance attenuation, and giving temperature early warning and positioning information according to the three-dimensional temperature field grid data. According to the invention, accurate positioning and real-time early warning of temperature abnormity at the early stage of the fire disaster are realized, and the accuracy and response speed of fire disaster monitoring are improved.

The invention relates to the technical field of fire prevention, and discloses a high-and-low-voltage power distribution room fire prevention early warning system based on potassium ion aerosol fire extinguishing, which comprises a smoke fire detection module used for detecting fire related data inside and outside a high-and-low-voltage power distribution room; the fire related data comprises smoke data and temperature data; the comprehensive evaluation module is used for judging whether an adopted evaluation model is a temperature evaluation model or a smoke evaluation model according to the fire related data; the fireproof grade classification module is used for grading fireproof grades in the high and low voltage power distribution room according to the screened evaluation model; the fire prevention measure implementation module is used for starting the potassium ion aerial fog fire extinguishing equipment in the corresponding fire prevention subareas to extinguish fire according to different grades; the electrical characteristic monitoring module is used for monitoring the load current and arc signal intensity of electrical equipment in the high-low voltage power distribution room and the temperature change rate of key nodes in real time; according to the invention, multi-dimensional accurate monitoring and graded prevention and control of the fire risk of the high and low voltage power distribution room are realized.

The invention relates to the technical field of fire detection, and provides an intelligent sensing multifunctional integrated sensor for fire. The sensor realizes synchronous real-time monitoring of a plurality of risk characteristics of a target area through a plurality of sensing monitoring modules triggered by a weight-based composite event, and when any sensor monitoring module monitors that the target area has the risk characteristics and the corresponding characteristic value reaches a preset characteristic threshold value, the risk characteristics of the target area can be monitored on the basis of different weight mechanisms. The dynamic risk weight of the target area is calculated through the feature dynamic weights of different sensing monitoring modules, and risk dynamic grading early warning is carried out on the target area; in the process of monitoring the target area by the sensing monitoring module, the feature dynamic weight of the sensing monitoring module dynamically changes along with the change of the feature value of the monitored risk feature, so that the fault-tolerant capability of the target area during risk early warning is improved, and the problems of false alarm and missing alarm caused by the single monitoring function of the existing sensor are effectively solved.

The subject matter of the present invention includes a fully automatic, early detection fire suppression method. The fire suppression method implements MEMS technology combined with artificial intelligence (AI) and machine learning (ML). The fire suppression method includes real-time monitoring, fire detection sensors, and diagnostics algorithms. A uniquely integrated application of technologies provides for distinguishing between safe and dangerously destructive fire events and the instantaneous extinguishing of a dangerous fire at its inception.

Multispectral imaging and related techniques are provided to detect fire hazards and alerts that are false positive indicators of the fire hazards. The embodiments are directed to systems for identifying pre-fire hazards using thermal images, visual images, and different rules corresponding to temperature configuration levels. The embodiment are also directed to identifying false positive indicators of fire hazards using regions of interest in the thermal and visual images, identifying objects that have hot components which may generate a false positive indicator for a fire hazard, and identifying sun reflections which may generate a false positive indicator for the fire hazard.

The invention discloses a fire behavior detection method based on multi-modal large model multi-scale analysis and depth reasoning, and relates to the technical field of artificial intelligence safety monitoring. The method comprises the following steps: acquiring visible light video data, infrared image data and audio data in the same fire scene; respectively preprocessing the data to obtain processed data; inputting the processed data into a corresponding modal encoder in a pre-trained ImageBind multi-modal model for processing, and respectively outputting a feature corresponding to each modal; a gating fusion module is adopted to perform fusion calculation on the features corresponding to the modals, and unified scene representation is output; and carrying out multi-dimensional depth reasoning by adopting a depth reasoning module to obtain a comprehensive result of the multi-dimensional reasoning, and outputting a fire confidence coefficient score and an interpretability basis based on the comprehensive result of the multi-dimensional reasoning. According to the invention, the accuracy of fire detection can be improved.

The invention belongs to the field of building fire detection, relates to a deep learning and image recognition technology, is used for solving the problem that fire analysis accuracy and fire alarm timeliness cannot be considered at the same time in the prior art, and particularly relates to a building fire early-stage detection method and system based on deep learning image recognition. The detection platform comprises a target marking module, a verification analysis module, a time sequence analysis module and a database. The target marking module is used for performing target marking analysis on the building fire: recording the internal environment of the building through a high-definition camera, decomposing the recorded video into frame-by-frame analysis images, amplifying the analysis images, segmenting the analysis images into a plurality of analysis areas, and extracting fire feature vectors in the analysis areas; according to the method and the system, closed-loop judgment from feature extraction to risk verification is completed in a single-frame image processing period, and due to the fact that data transmission delay between modules is eliminated, the system can respond to potential fire behaviors in a millisecond level.

The invention discloses an integrated fire detection intelligent lamp for a railway tunnel. The integrated fire detection intelligent lamp comprises a multi-mode sensing module, an integrated thermal imaging array sensor, a multispectral flame sensor, a laser scattering type smoke sensor and an electrochemical CO sensor. An FPGA chip and a lightweight convolutional neural network model are arranged in the signal processing module, space-time fusion analysis is carried out on multi-modal sensing data in real time, and a fire judgment threshold is dynamically optimized through transfer learning; the dynamic ad hoc network communication module supports dual protocols of LoRaWAN and TSN; the adaptive power supply adopts an AC / DC power conversion module, integrates a piezoelectric vibration energy collection module and a thermoelectric power generation module, and combines a super capacitor and a lithium battery to realize quintuple redundant power supply; and the three-dimensional lighting array consists of a plurality of groups of independently controlled LED matrixes and is used for dynamically adjusting a lighting area and brightness distribution. The fire hazard detection sensitivity is improved through multi-dimensional data fusion, the illumination function is achieved, and the use requirement of the railway tunnel in the severe environment is met.

The invention relates to a fire area inversion method based on airborne dual-spectrum detection and depth estimation, and belongs to the field of unmanned aerial vehicle detection. The method comprises the following steps: acquiring a multi-dimensional data set of a dual-spectrum image, a temperature image, an unmanned aerial vehicle attitude and the like; constructing a multi-modal space collaborative perception segmentation network, combining temperature change characteristics with temperature space distribution characteristics of flames to generate temperature region distribution characteristics, and coupling absolute temperature and pixel information to enhance flame weak edge extraction; designing a temperature-guided space structure loss function TSSLoss, and combining gradient change consistency constraint and temperature weight constraint on a segmentation loss function for network training; and according to the unmanned aerial vehicle pose, the target depth and the fire area segmentation pixel area, an early fire area is derived in combination with an airspace transmission inversion formula, and an actual fire area is calculated. According to the method, multi-source information is effectively combined for physical constraint, and more accurate fire detection segmentation and fire area calculation can be realized.

The invention relates to the technical field of fire detection, in particular to a smog alarm method based on a smog particle motion trail, which comprises the following steps: constructing a three-dimensional flow field distribution diagram of smog particles, and aiming at the sectional area change and curvature change of a current channel at a branch path, taking the predicted smog particle position as a center to obtain a smog particle motion trail distribution diagram; associating the flow velocity data of each branch path to obtain a flow velocity gradient sequence of each branch path; on the basis of the flow velocity gradient sequence of each branch path, when the flow velocity gradient exceeds a flow velocity threshold value, the smoke particle position in each branch path is updated; performing trajectory fitting on the smoke particles, and configuring concentration intensity indexes and updating time of the smoke particles in different branch paths according to concentration gradients corresponding to the motion trajectory sequences on the branch paths; and configuring a classification label of each branch path, and carrying out graded early warning according to flow velocity differentiation. And the accuracy and timeliness of smoke alarm are improved.

The utility model provides a fire safety monitoring system, and belongs to the technical field of fire protection. The fire safety monitoring system comprises a smoke detection module, a temperature detection module, an infrared thermal imaging module, a control module, a switch module and a fire extinguishing module, the smoke detection module, the temperature detection module and the infrared thermal imaging module are all connected with the control module; the switch module is connected with the control module and the fire extinguishing module. The smoke detection module is configured to detect the smoke concentration in the environment; the temperature detection module is configured to detect the temperature in the environment; the infrared thermal imaging module is configured to position a fire source; the control module is configured to receive detection information of the smoke detection module, the temperature detection module and the infrared thermal imaging module and control the fire extinguishing module according to the detection information. The fire detection precision can be improved.

The invention relates to the technical field of fire detection, in particular to a photovoltaic power station outdoor fireproof monitoring method and system, and the method comprises the following steps: obtaining a multi-source time record, screening a synchronous window, extracting gray scale distribution, forming a connected heat domain fragment, generating a survival span structure according to a gray scale sequence, and recognizing the position of a heat source based on coordinate aggregation. And generating a fire source correlation judgment structure in combination with thermal change and electrical change trends. According to the method, the cross-source synchronous window is established in the time dimension to realize matching of multi-source data in a continuous interval, so that thermal image change and electrical change are analyzed under the same time reference, and thermal domain connectivity, coverage form and gray evolutionary structure are refined in the space dimension, so that the thermal image change and the electrical change are analyzed. By introducing time consistency analysis of thermal characteristics and electrical increment trends into correlation judgment, a monitoring result is changed from static threshold judgment to dynamic recognition based on time-space consistency, and the accuracy and reliability of photovoltaic scene fire source recognition are remarkably improved.

The invention discloses a wind power cabin fire early warning device and a wind power cabin fire early warning method based on multispectral fusion, and aims at the defects of high false alarm rate, insufficient motion artifact compensation and transient fire capture lag of single-spectrum detection in the prior art to construct a multispectral collaborative analysis architecture. A thermal deformation dynamic compensation model is constructed in combination with a main shaft rotating speed signal, and sub-pixel-level motion artifact elimination is achieved; an improved three-dimensional convolutional neural network is adopted to extract smoke diffusion trajectory, temperature conduction trend and arc pulse features, and multi-source feature adaptive fusion is realized through a cross-modal attention mechanism and a graph neural network; and establishing an oil stain reflection spectrum fingerprint database for false alarm filtering, and finally triggering third-level early warning according to an environment self-adaptive threshold value. According to the method, the electrical short circuit detection time can be greatly shortened, the false alarm rate is effectively reduced, the smoldering fire is warned in advance, and the industrial problem of accurate recognition of early fire in a complex cabin environment is solved.

The invention discloses an intelligent tunnel fire identification method and system based on multi-modal sensor fusion, particularly relates to the field of tunnel safety monitoring, and is used for solving the problems of fire detection delay, inaccurate positioning and low response efficiency in an ultra-long tunnel. Fusing and compressing into a global identification quantity delineated abnormal region; then, activating the monorail suspension device by taking the suspected area as a target, and generating a local recognition quantity by a second sensor; forming a corrected recognition set through time-space alignment and deviation correction; on the basis, a fire source positioning result set is extracted, temperature deviation intensity and a smoke propagation rate coefficient are analyzed through machine learning, and a fire development potential index is calculated; and finally, writing back a result to correct a trigger parameter and a travel path, and supporting subsequent fire reuse. The recognition precision is improved, the response time is shortened, and casualty and loss risks are reduced.

The invention discloses a fire-fighting detection early warning system, and relates to the technical field of fire-fighting early warning, and the system comprises a multi-source data obtaining module which comprises a gas monitoring unit, a smoke monitoring unit and a temperature obtaining unit, obtains the concentration of combustible gas, the granularity of smoke and the temperature data, carries out the preprocessing of the data, and integrates the data into a data set; the analysis and early warning module analyzes the data based on the data set, introduces a gradient compensation algorithm into the temperature acquisition unit, synchronously records the ambient temperatures of three different distances around, calculates a temperature gradient value, judges whether the temperature data exceeds the standard or not based on the gradient value, preliminarily obtains a judgment result from the temperature direction, and sends the judgment result to the early warning unit; and performing trend prediction on the multi-source data based on a time sequence analysis method. According to the method, the gradient compensation algorithm is introduced, the gradient value is calculated by using the three temperature monitoring points at different distances, local temperature anomaly can be accurately captured, the danger diffusion trend can be recognized earlier, local fluctuation and system risks can be distinguished, and time is won for early intervention.

The embodiment of the invention provides a smoke shielding false alarm intelligent identification system and method, and is applied to the technical field of intelligent fire detection, and the method comprises the steps: synchronously collecting an imaging picture, radiant heat response data and a wind field state parameter of a suspected shielding region; the method comprises the following steps: identifying and identifying an abnormal shielding area in a picture; further analyzing the motion characteristics of the region along with the change of the wind field and the dynamic process of the thermal inertia change of the region; the core is to judge whether a time sequence lag relationship meeting a smoke diffusion physical rule exists between the two, so as to identify whether the shielding object is combustion smoke or not. And if the physical mechanism is not met, the alarm is inhibited, so that real fire smoke and non-fire interferents such as dust and steam are effectively distinguished, and the fire false alarm rate in a complex ventilation environment is remarkably reduced.

The invention relates to the technical field of aviation safety monitoring, in particular to an aircraft cargo hold fire alarm detection system based on AI anomaly recognition and thermal imaging linkage, and the system comprises a visible light image collection module which is used for obtaining a real-time visible light image in an aircraft cargo hold; the infrared thermal imaging acquisition module is used for acquiring multi-region and multi-channel thermal imaging images in the cargo hold and constructing a thermal distribution diagram; the smoke and gas sensing device is used for monitoring the concentration change of hazardous gas in the air and outputting continuous detection data; the multi-modal AI identification processing module is used for carrying out fusion analysis on the multi-modal data; the pod linkage verification module is used for performing external verification sampling on the target area when the multi-modal analysis result is low in confidence but has risks, and forming a final judgment result in an auxiliary manner; and the alarm output module is used for converting the final judgment result into an alarm signal, sending the alarm signal to a cockpit HUD or a main control system through a flight control system, and automatically switching a prompt mode according to the fire behavior grade.

The invention discloses a control system of a device for drying metal powder containing combustible volatile substances, and belongs to the technical field of automatic control. In the system, a first subtracter generates a sequence first error signal according to the weight of two pieces of metal powder provided by a weighing sensor at an interval of a time step length; the second subtracter generates a sequence second error signal according to the actually measured temperature and the set temperature in the drying cavity provided by the temperature sensor; the reinforcement learning module respectively generates a first state and a second state according to the sequence first error signal and the sequence second error signal, and generates a control strategy of a PID controller for controlling the working state of the heater according to the first state and the second state; and the fire detection model estimates the probability value of the fire in the drying cavity according to the oxygen concentration in the drying cavity, the concentration of the volatile matter and the actually measured temperature, and if the probability value exceeds a threshold value, the stored inert gas is injected into the drying cavity to reduce the probability of the fire. The explosion risk caused by combustible gas accumulation is avoided.

The invention relates to the technical field of fire detection and early warning, in particular to an intelligent ultraviolet flame detection and early warning system and method suitable for complex industrial scenes. Comprising distributed multispectral ultraviolet detection nodes and a central processing early warning unit. The nodes carry out signal acquisition and preliminary identification through a multispectral sensor, local processing and high-precision time synchronization; and the central unit fuses multi-node data, performs three-dimensional flame positioning by using a TDOA algorithm, performs machine learning recognition and false alarm elimination, and performs trajectory tracking by using a Kalman filter so as to realize graded early warning linkage. According to the intelligent ultraviolet flame detection and early warning system and method suitable for the complex industrial scene, the problems that in the prior art, the false alarm rate is high, flame positioning is not accurate, trajectory tracking is missing, cooperative monitoring is limited and the like are solved or at least relieved, false alarms are effectively reduced, and the positioning precision, the tracking capacity and the early warning reliability are improved.

The invention discloses an image adaptive feature enhancement and layered attention method for forest fire detection, and belongs to the technical field of computer vision and forest fire prevention. According to the scheme, on the basis of a DEIM framework, a self-collection data set containing multiple scenes is constructed and preprocessed; after collected images are standardized, features are extracted through a multi-scale bidirectional feature transfer backbone network, the features are refined and enhanced through a feature fusion and diffusion module, then a MetaFormer architecture hierarchical attention encoder (including SHSA and EDFFN) is used for encoding, and a fire behavior is judged in combination with Dense-O2O matching (three types of fire behavior templates) and an F1 maximized adaptive threshold value. The method realizes that the mAP at 50 of a small target reaches 76.5%, the mAP at 50 of a complex scene reaches 77.8%, the parameter quantity is 3.3 M, edge equipment reasoning is 28.2 FPS, false detection and missing detection are balanced, and the method adapts to edge deployment of an unmanned aerial vehicle and the like.