Intelligent fire extinguishing method and system and fire fighting equipment

Abstract

The embodiment of the invention provides an intelligent fire extinguishing method and system and fire fighting equipment, and belongs to the technical field of intelligent fire fighting. The method is realized based on linkage control of an arm support and a fire monitor of fire fighting equipment. The method comprises the following steps: calling a corresponding fire source identification model based on multi-dimensional scanning information; wherein the fire source identification model is obtained by training based on training sample data containing fire source multi-dimensional features; when the fire source recognition model recognizes a fire source, judging the position relation between the position of the fire source and the position of the fire-fighting equipment; on the basis of the judgment result of the position relation, the position and posture adjustment of the fire-fighting equipment is executed until the fire-fighting equipment reaches a fire extinguishing position and posture state; wherein the position and posture of the fire-fighting equipment comprises position information of the fire-fighting equipment, posture information of an arm support and posture information of a fire monitor. According to the scheme, the accuracy of fire source recognition and the fire extinguishing efficiency of fire fighting equipment are remarkably improved, meanwhile, the higher adaptive capacity is achieved in the complex fire scene environment, and an efficient solution is provided for intelligent fire extinguishing.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent fire fighting, and specifically relates to an intelligent fire extinguishing method, an intelligent fire extinguishing system and a fire fighting device. Background Art

[0002] Fire fighting devices (such as fire trucks and fire fighting robots), as important equipment for fire fighting, their intelligent level has a crucial impact on fire extinguishing efficiency and safety. In the prior art, the intelligent fire fighting cannon control system has achieved a certain degree of automation, and the fire source can be detected and tracked through machine vision and AI technology. However, most of these technologies focus on the control of a single fire fighting cannon, ignoring the necessity of the coordinated operation of the boom and the muzzle of the fire fighting device. In practical applications, the complex fire scene environment requires the boom of the fire fighting device to provide more flexible adjustment to expand the shooting range and coverage area, but the prior art fails to achieve the linkage control of the boom and the muzzle, resulting in the difficulty of achieving the expected fire source tracking and fire extinguishing effect.

[0003] The accuracy of fire source detection and environmental adaptability have also become bottlenecks in the prior art. Traditional fire source detection models usually rely on single visual or infrared features, lacking the deep integration of the temperature characteristics, dynamic changes and shape characteristics of the fire source. In the dynamically changing fire scene environment, this technical deficiency is likely to lead to fire source positioning errors and reduce the accuracy of fire extinguishing. At the same time, lidar detection, the core of the jet control system, also faces the problem of blocked vision. The traditional single-sided lidar layout is prone to occlusion under the interference of the jet or crosswind, and cannot accurately detect the position of the jet landing point. This not only affects the control accuracy of the jet, but also makes it impossible to efficiently implement the fire extinguishing closed-loop control.

[0004] Therefore, the core problems of the prior art are the failure to achieve the linkage cooperation between the boom and the muzzle of the fire fighting device, and the lack of accurate detection methods for the dynamic characteristics of the fire source and the jet landing point. These problems lead to the slow and inaccurate response of the fire extinguishing system in the complex fire scene environment, making it difficult to meet the high efficiency and automation requirements of modern fire fighting. Summary of the Invention

[0005] The purpose of the embodiments of the present invention is to provide an intelligent fire extinguishing method, system and a fire fighting device, so as to at least solve the problems that the prior art fails to achieve the linkage cooperation between the boom and the muzzle of the fire fighting device, and lacks accurate detection methods for the dynamic characteristics of the fire source and the jet landing point.

[0006] To achieve the above object, a first aspect of the present invention provides an intelligent fire extinguishing method, which is realized based on the linkage control of the boom and the fire monitor of the fire fighting equipment. The method includes: scanning for a fire source within a predetermined visible range to obtain real-time multi-dimensional scanning information, and calling a corresponding fire source recognition model based on the multi-dimensional scanning information; wherein, the fire source recognition model is trained based on training sample data including multi-dimensional features of the fire source, and the multi-dimensional features of the fire source include any one or more of the fire source temperature information, the dynamic characteristics of the fire source, and the fire source position information; when the fire source recognition model recognizes a fire source, determining the positional relationship between the fire source position and the position of the fire fighting equipment; based on the result of the positional relationship determination, performing pose adjustment of the fire fighting equipment until the fire fighting equipment reaches the fire extinguishing pose state; wherein, the pose of the fire fighting equipment includes the position information of the fire fighting equipment, the pose information of the boom, and the pose information of the fire monitor.

[0007] Optionally, a sensor is provided on the nozzle end side of the fire monitor; the sensor includes any one or more of a visible light camera, an infrared camera, and a lidar; the multi-dimensional scanning information includes any one or more of video stream information, infrared image information, and radar response information.

[0008] Optionally, the method further includes performing training of the fire source recognition model; the training rule of the fire source recognition model is: collecting historical image information, performing multi-dimensional feature annotation of the fire source in the historical image information, and constructing a training sample based on the annotated images; performing model training on a pre-constructed neural network based on the training set in the training sample, and updating the network weights through backpropagation and optimization algorithms during the training process to minimize the loss function to obtain an initial model; performing verification on the initial model based on the validation set in the training sample, and using the model that passes the verification as the fire source recognition model.

[0009] Optionally, the acquisition rule of the dynamic characteristics of the fire source is: performing pixel displacement recognition between consecutive image frames based on the optical flow method, and obtaining the dynamic characteristics of the fire source based on the recognition result.

[0010] Optionally, the construction rule of the training sample is: performing feature combination on each fire source feature in the multi-dimensional features of the fire source based on a single historical image information to obtain a set of combined features corresponding to one historical image information; constructing a corresponding training sample based on the combined features of all historical image information.

[0011] Optionally, the pre-built neural network is constructed based on the YOLO network, and the pre-built neural network includes: a combined feature input layer for performing the input of training samples of combined features; a convolutional feature extraction layer including a fire source temperature feature extraction layer and a fire source dynamic feature extraction layer for performing convolutional processing on the input combined features; an attention mechanism module including an SE module or a CBAM module for enhancing the attention of features; and a loss calculation module for determining the total loss based on each loss function.

[0012] Optionally, each loss function includes any one or more of: a classification loss function, a localization loss function, a temperature loss function, and a jump loss function; the classification loss function is used to determine whether the target area contains a flame; the localization loss function is used to optimize the coordinates and dimensions of the flame bounding box; the temperature loss function is used to predict the temperature distribution in the flame area and align it with the true value; the jump loss function is used to determine the change in the dynamic features of the flame and optimize the prediction of dynamic information.

[0013] Optionally, when the fire source recognition model recognizes a fire source, the determination of the positional relationship between the fire source position and the fire fighting equipment position is performed, including: when the fire source recognition model recognizes a fire source, the three-dimensional position coordinates of the target fire source are determined based on the combination of the visual image information and the radar response information of the target fire source in the multi-dimensional scan information; based on the pose information of the fire fighting equipment, the three-dimensional position coordinates of the target fire source, and the fire fighting equipment range function of the fire fighting equipment, it is determined whether the target fire source is within the range of the fire fighting equipment, and the determination result is used as the determination result of the positional relationship between the fire source position and the fire fighting equipment position.

[0014] Optionally, the fire fighting equipment range function is constructed based on the range calculation model of the Bernoulli equation; the influencing factors of the fire fighting equipment range function include: the pose information of the fire fighting equipment, the performance parameter information of the fire fighting cannon, and the environmental parameter information.

[0015] Optionally, based on the result of the determination of the positional relationship, the pose adjustment of the fire fighting equipment is performed until the fire fighting equipment reaches the fire extinguishing pose state, including: if the determination result of the positional relationship between the fire source position and the fire fighting equipment position is that the target fire source is within the range of the fire fighting equipment, then the result signal that the fire fighting equipment finally reaches the fire extinguishing pose state is directly output based on the current pose state of the fire fighting equipment; if the determination result of the positional relationship between the fire source position and the fire fighting equipment position is that the target fire source is not within the range of the fire fighting equipment, then the pose adjustment of the fire fighting equipment is performed based on the inverse kinematics algorithm until the fire fighting equipment reaches the fire extinguishing pose state.

[0016] Optionally, performing the pose adjustment of the fire-fighting equipment based on the inverse kinematics algorithm until the fire-fighting equipment reaches the fire-extinguishing pose state includes: determining the target pose information of the fire-fighting equipment within the range of the fire-fighting equipment based on the three-dimensional position coordinates of the target fire source and the range function of the fire-fighting equipment; performing the inverse kinematics algorithm based on the current pose information of the fire-fighting equipment and the target pose information to obtain multiple planned paths from the current pose information of the fire-fighting equipment to the target pose information; performing feasibility verification on each planned path based on kinematic constraints, filtering out the planned paths that cannot meet the kinematic constraints, and taking the planned path with the shortest time-consuming in the remaining planned paths as the target path; performing the pose adjustment of the fire-fighting equipment based on the target path until the fire-fighting equipment reaches the fire-extinguishing pose state.

[0017] Optionally, the method further includes: during the process of performing the pose adjustment of the fire-fighting equipment, performing real-time scanning of the target fire source; when the three-dimensional position coordinates of the target fire source change, updating the target pose information of the fire-fighting equipment within the range of the fire-fighting equipment based on the new three-dimensional position of the target fire source, and updating the target path based on the updated target pose information; performing the pose adjustment of the fire-fighting equipment based on the updated target path until the fire-fighting equipment reaches the fire-extinguishing pose state.

[0018] Optionally, a plurality of lidars are arranged in different directions on the nozzle end side of the fire cannon; the determination rule of the radar response information is: performing synchronous scanning based on each lidar to respectively obtain the three-dimensional point cloud data of the jet area of each lidar; performing spatial registration based on the three-dimensional point cloud data of the jet area of each lidar; performing frame difference method operation on the lidar data before and after the fire cannon sprays water based on each lidar after completing spatial registration to identify the jet area; performing fitting on the identified jet area based on the RANSAC algorithm to obtain the fitted jet area as the radar response information.

[0019] The second aspect of the present invention provides an intelligent fire extinguishing system, which is realized based on the linkage control of the boom and the fire cannon of the fire-fighting equipment. The system includes: a scanning unit, configured to perform fire source scanning within a predetermined visible range to obtain real-time multi-dimensional scanning information, and call a corresponding fire source recognition model based on the multi-dimensional scanning information; wherein, the fire source recognition model is trained based on training sample data including multi-dimensional features of the fire source, and the multi-dimensional features of the fire source include any one or more of fire source temperature information, fire source dynamic features, and fire source position information; a position determination unit, configured to perform a position relationship determination between the fire source position and the position of the fire-fighting equipment when the fire source recognition model recognizes a fire source; a pose adjustment unit, configured to perform the pose adjustment of the fire-fighting equipment based on the result of the position relationship determination until the fire-fighting equipment reaches the fire-extinguishing pose state; wherein, the pose of the fire-fighting equipment includes the position information of the fire-fighting equipment, the attitude information of the boom, and the attitude information of the fire cannon.

[0020] Optionally, a sensor is provided on the nozzle end side of the fire monitor; the sensor includes any one or more of a visible light camera, an infrared camera, and a lidar; a plurality of lidars are provided in different directions along the nozzle end side of the fire monitor.

[0021] The third aspect of the present invention provides a fire fighting device, and the fire fighting device is configured with the above-mentioned intelligent fire extinguishing system.

[0022] On the other hand, the present invention provides a computer-readable storage medium, and instructions are stored on the computer-readable storage medium, and when the instructions are run on a computer, the computer is caused to execute the above-mentioned intelligent fire extinguishing method.

[0023] Through the above technical solutions, the solution of the present invention scans for a fire source within a predetermined visible range, obtains real-time multi-dimensional scan information, and uses an identification model trained based on multi-dimensional features of the fire source to achieve accurate identification of the fire source. The fire source identification model improves the accuracy of fire source detection and environmental adaptability by fusing multi-dimensional features such as temperature and dynamic characteristics. When a fire source is detected, the system further determines the positional relationship between the fire source and the fire fighting device, and combines the current pose and the firing range of the fire fighting device to dynamically plan the optimal fire extinguishing pose of the fire fighting device. By automatically adjusting the pose of the fire fighting device, it is ensured that the fire source is within the effective firing range of the fire monitor and quickly enters the fire extinguishing state. The solution of the present invention significantly improves the accuracy of fire source identification and the fire extinguishing efficiency of the fire fighting device, and at the same time has stronger adaptability in a complex fire field environment, providing an efficient solution for intelligent fire extinguishing.

[0024] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent specific embodiment part. Description of the Drawings

[0025] The drawings are used to provide a further understanding of the embodiments of the present invention, and constitute a part of the specification, and are used together with the following specific embodiments to explain the embodiments of the present invention, but do not constitute a limitation to the embodiments of the present invention. In the drawings:

[0026] Figure 1 is a flowchart of the steps of an intelligent fire extinguishing method provided by an embodiment of the present invention;

[0027] Figure 2 is a schematic structural diagram of the boom and the fire monitor of a fire fighting device provided by an embodiment of the present invention;

[0028] Figure 3 is a system structure diagram of an intelligent fire extinguishing system provided by an embodiment of the present invention. Detailed Embodiments

[0029] The following will describe in detail the specific embodiments of the present invention with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only for the purpose of illustrating and explaining the present invention, and are not intended to limit the present invention.

[0030] Figure 1 is a flowchart of a method for an intelligent fire extinguishing method provided by an embodiment of the present invention. As Figure 1 shown, an embodiment of the present invention provides an intelligent fire extinguishing method, which is realized based on the linkage control of the boom and the fire monitor of the fire fighting equipment.

[0031] First of all, the solution of the present invention is applicable to fire fighting equipment with a boom and a fire monitor, and has broad application prospects especially in equipment such as fire trucks and fire fighting robots. These devices are usually used for efficient fire source extinguishing and rescue operations at the fire scene. The boom can flexibly adjust the angle and height to ensure that the fire monitor can accurately aim at the fire source and spray high-pressure water or fire extinguishing agent. As an important part of the modern fire fighting system, fire trucks and fire fighting robots can be remotely controlled and efficiently extinguish fires in complex and dangerous environments, reducing manual participation, thereby improving the fire extinguishing efficiency and ensuring the safety of rescue personnel. In addition, the solution of the present invention can also be extended to other types of fire fighting equipment, such as automated fire extinguishing systems, unmanned fire trucks, etc., and has wide applicability.

[0032] Specifically, the method includes:

[0033] Step S10: Scan the fire source within a predetermined visible range to obtain real-time multi-dimensional scan information, and call the corresponding fire source recognition model based on the multi-dimensional scan information.

[0034] Specifically, as Figure 2 , a sensor is provided on the head end side of the fire monitor; the sensor includes any one or more of a visible light camera, an infrared camera, and a lidar; the multi-dimensional scan information includes any one or more of video stream information, infrared image information, and radar response information.

[0035] Furthermore, a visible light camera is used to obtain real-time video stream information of the fire scene, providing clear images of the fire source and the environmental background; an infrared camera obtains temperature distribution information of the fire source by detecting the thermal radiation of the fire scene, and is particularly suitable for smoke-obscured or low-light environments; a lidar senses the distance, position of the fire source and the spatial distribution of the jet landing points in real time through high-precision three-dimensional point cloud data acquisition. The multi-dimensional scanning information generated by these three sensors, including video stream information, infrared image information and radar response information, can provide comprehensive and accurate data support for fire source detection and pose adjustment. By fusing this multi-dimensional information, a fire source recognition model can be used to extract multi-dimensional features of the fire source, significantly improving the accuracy and robustness of fire source detection. At the same time, the three-dimensional point cloud data generated by the lidar can also be combined with the frame difference method and the RANSAC algorithm to dynamically monitor the jet area, identify the jet landing points, and match them with the fire source position to achieve closed-loop fire extinguishing control.

[0036] Preferably, the method further includes performing training of the fire source recognition model; the training rule of the fire source recognition model is: collecting historical image information, performing multi-dimensional feature annotation of the fire source in the historical image information, and constructing a training sample based on the annotated images; performing model training on a pre-constructed neural network based on the training set in the training sample, and updating the network weights through backpropagation and optimization algorithms during the training process to minimize the loss function and obtain an initial model; performing verification on the initial model based on the validation set in the training sample, and using the model that passes the verification as the fire source recognition model.

[0037] Furthermore, the multi-dimensional features of the fire source include any one or more of fire source temperature information, fire source dynamic features and fire source position information; the acquisition rule of the fire source dynamic features is: performing pixel displacement recognition between consecutive image frames based on the optical flow method, and obtaining the dynamic features of the fire source based on the recognition result.

[0038] Specifically, the construction rule of the training sample is: performing feature combination on each fire source feature in the multi-dimensional features of the fire source based on a single historical image information to obtain a set of combined features corresponding to one historical image information; constructing corresponding training samples based on the combined features of all historical image information.

[0039] In the embodiment of the present invention, a large amount of historical image information is collected, and these images cover different scenes, lighting conditions and fire source characteristics to ensure data diversity and representativeness. In these images, the multi-dimensional features of the fire source are accurately annotated, including the temperature information, dynamic features and position information of the fire source, etc., to ensure the comprehensiveness and high quality of the training data. Based on these annotated images, a training sample set including a training set and a validation set is constructed to provide data support for model training.

[0040] During the model training phase, the training set data is input into a pre - constructed neural network, and deep learning methods are used for feature extraction and pattern learning. In particular, the neural network architecture is based on the YOLO network and includes a combined feature input layer, a convolutional feature extraction layer, and an attention mechanism module. During the training process, the network weights are updated through the backpropagation algorithm and optimization algorithms (such as Adam or SGD), gradually reducing the prediction error to minimize the loss function. The loss function design not only includes classification loss and localization loss, but also incorporates temperature loss and dynamic feature loss to comprehensively optimize the model performance. After training is completed, the model will be verified in the validation set to ensure its accuracy and generalization ability on unseen data. Only the model that passes the verification will be determined as the final fire source recognition model.

[0041] In the embodiment of the present invention, during the training process of the deep learning model, the backpropagation algorithm and the optimization algorithm work together to update the model weights, helping the model gradually approach the optimal solution. Among them, the backpropagation algorithm is the most commonly used error correction algorithm in neural networks. It calculates the gradient of the loss function with respect to each weight through the chain rule and backpropagates these gradients to each layer of the network, thereby updating the weight parameters. Its key idea is to calculate the error step by step from the output layer to the input layer and adjust the weights of each layer to minimize the loss function.

[0042] Optimization algorithms (such as Adam and SGD) are the execution tools for backpropagation, which determine how to update the network weights based on the gradient information. Specifically, Stochastic Gradient Descent (SGD) is the most basic optimization algorithm. It calculates the gradient by using a small batch of data each time to update the model parameters. Although the SGD algorithm is simple and easy to implement, its convergence speed is slow, and it is easily affected by the learning rate setting and local minima. The Adam (Adaptive Moment Estimation) optimizer, by combining the momentum of the gradient and the adaptive adjustment of the gradient (adaptive learning rates), can automatically adjust the learning rate of each parameter, significantly improving the training stability and convergence speed. This adaptive feature of Adam is especially suitable for complex deep learning networks, which can accelerate the training process while reducing manual adjustment of the learning rate.

[0043] Based on the solution of the present invention, compared with the traditional gradient descent method, the Adam optimizer has a faster convergence speed and stronger robustness. Especially when dealing with highly nonlinear and high-dimensional tasks, it can effectively avoid the problems of gradient vanishing or gradient explosion. Through the combination of these optimization algorithms, the backpropagation algorithm demonstrates its important advantages in model training. Especially when facing complex tasks, such as object recognition problems in the YOLO network structure, it can quickly and effectively adjust the network parameters, improving the training efficiency and the final generalization ability of the model. Therefore, in the training of the fire source recognition model, the combination of backpropagation and the Adam optimization algorithm not only accelerates the training process but also effectively improves the recognition accuracy. Especially when dealing with large-scale datasets and complex features, it can better find the patterns in the data and make accurate predictions.

[0044] Furthermore, the dynamic features of the fire source are obtained by the optical flow method. Between consecutive image frames, the optical flow method is used to analyze the displacement changes of pixels to extract the dynamic information of the fire source. The optical flow method can capture the motion characteristics of the fire source, especially in a rapidly changing fire scene environment, and this characteristic is crucial for accurately detecting the position of the fire source. The temperature information is extracted by an infrared thermal imaging camera, which can accurately locate the high-temperature area, that is, the center point of the fire source. The position information is combined with lidar data to provide the three-dimensional coordinates of the fire source. Through the fusion of these features, the model has higher robustness and adaptability.

[0045] In terms of the rules for constructing training samples, in order to maximize the utilization rate of feature information, the multi-dimensional features (such as temperature, dynamic features, position, etc.) in a single historical image information are combined to form a complete combined feature. The information combined features of all historical images are uniformly constructed as the training samples of the model. This method of feature combination can enrich the diversity of training data and at the same time help the model learn the correlation between different features.

[0046] Specifically, the pre-constructed neural network is constructed based on the YOLO network, and the pre-constructed neural network includes: a combined feature input layer for performing the input of the training samples of the combined features; a convolutional feature extraction layer, including a fire source temperature feature extraction layer and a fire source dynamic feature extraction layer, for performing convolutional processing on the input combined features; an attention mechanism module, including an SE module or a CBAM module, for enhancing the attention to the features; and a loss calculation module for determining the total loss based on each loss function.

[0047] Furthermore, each loss function includes any one or more of a classification loss function, a localization loss function, a temperature loss function, and a jump loss function; the classification loss function is used to determine whether the target area contains a flame; the localization loss function is used to optimize the coordinates and size of the flame bounding box; the temperature loss function is used to predict the temperature distribution of the flame area and align it with the true value; the jump loss function is used to determine the change in the dynamic characteristics of the flame and optimize the prediction of dynamic information.

[0048] In an embodiment of the present invention, the pre-constructed neural network is constructed based on the YOLO network (such as YOLOv8), optimized for the fire source detection task, and integrates multiple modules to improve the detection accuracy and adaptability of the model.

[0049] Furthermore, the combined feature input layer is responsible for inputting multi-dimensional fire source features into the network. These features include the temperature information, dynamic features, and position information of the fire source, enhancing the richness of the input data through feature combination. Compared with traditional single input, this multi-channel feature input method can more comprehensively capture the physical and dynamic properties of the fire source, laying a foundation for subsequent feature extraction. The convolutional feature extraction layer extracts features from the input multi-dimensional features through a deep convolutional network, including a temperature feature extraction layer and a dynamic feature extraction layer specifically for the fire source. The temperature feature extraction layer uses the infrared image information to extract the characteristics of the high-temperature area and accurately locates the center of the fire source, while the dynamic feature extraction layer processes the pixel changes in consecutive frames through the optical flow method to capture the dynamic behavior and shape changes of the fire source. This two-channel feature extraction design can effectively improve the robustness of the model in complex scenarios, especially having significant advantages in a rapidly changing fire scene environment.

[0050] Furthermore, to further enhance the expression ability of features, the network also introduces an attention mechanism module, including the SE (Squeeze-and-Excitation) module and the CBAM (Channel Attention and Spatial Attention) module. These modules significantly enhance the model's ability to focus on the key areas of the fire source by dynamically adjusting the weights of feature channels and spatial positions. For example, the SE module generates the weight distribution of each feature channel through global average pooling, emphasizing the channel information related to the fire source; the CBAM module combines channel attention and spatial attention to identify the position of the fire source and its changes in a more refined way. This attention enhancement mechanism effectively reduces the interference of background noise and improves the detection accuracy.

[0051] Furthermore, the loss calculation module integrates multiple loss functions for comprehensively evaluating different objectives in the fire source detection task. The loss functions include classification loss, localization loss, temperature loss, and jump loss. Specifically, it includes:

[0052] 1) The classification loss maintains the native classification loss of YOLOv8 to determine whether the target area contains fire. By optimizing the classification results, it ensures that the fire source can be accurately identified. Specifically, it includes:

[0053] 2) The localization loss optimizes the coordinates and dimensions of the fire bounding box to ensure the accuracy of fire source localization, expressed as:

[0054] L = L box + λ1L temp + λ2L change

[0055] where L box is the basic bounding box regression loss; L temp and L change are the regression losses corresponding to temperature and jump information respectively; λ1 and λ2 are balancing weights.

[0056] 3) The temperature loss further optimizes the detection accuracy of the high-temperature area by predicting the temperature distribution in the fire source area and aligning it with the true value, expressed as L temp = (P - T) 2 .

[0057] 4) The jump loss focuses on the changes in dynamic features. By optimizing the prediction of fire source movement and shape changes, it improves the adaptability of the model in dynamic environments. Through the weighted combination of these loss functions, the network can achieve global optimization in multiple aspects such as classification, localization, temperature detection, and dynamic feature analysis, forming a complete fire source detection system, expressed as L change = ||P change - M change || 2 .

[0058] The design of this neural network fully integrates multi-dimensional fire source features, significantly improving the detection accuracy and robustness. The combined feature input and dedicated feature extraction layers enable the network to extract key features from multi-dimensional information such as temperature, dynamic characteristics, and location, solving the adaptability problem of traditional fire source detection methods in complex environments. The attention mechanism module further enhances the ability to focus on key areas of the fire source, reducing background noise interference. The multi-dimensional optimization strategy of the loss calculation module ensures the comprehensiveness and accuracy of fire source detection. The overall network structure provides an efficient and reliable technical foundation for intelligent fire protection, especially suitable for fire source detection and fire extinguishing tasks in complex dynamic fire scenes, achieving a comprehensive optimization of fire source identification and localization, and providing more powerful technical support for precise fire extinguishing.

[0059] Step S20: When the fire source recognition model recognizes a fire source, determine the positional relationship between the fire source location and the location of the fire fighting equipment.

[0060] Specifically, when the fire source recognition model recognizes a fire source, the three-dimensional position coordinates of the target fire source are determined based on the combination of the visual image information and the radar response information of the target fire source in the multi-dimensional scanning information; based on the pose information of the fire fighting equipment, the three-dimensional position coordinates of the target fire source, and the fire fighting equipment range function of the fire fighting equipment, it is determined whether the target fire source is within the range of the fire fighting equipment, and the judgment result is used as the judgment result of the position relationship between the fire source position and the fire fighting equipment position; wherein, the pose information of the fire fighting equipment includes the three-dimensional position information of the fire fighting equipment and the attitude information of the fire fighting equipment; the attitude information of the fire fighting equipment includes the muzzle angle information of the fire fighting cannon and / or the joint parameter information of the boom.

[0061] Further, the fire fighting equipment range function is constructed based on the range calculation model of the Bernoulli equation; the influencing factors of the fire fighting equipment range function include: the attitude information of the fire fighting equipment, the performance parameter information of the fire fighting cannon, and the environmental parameter information.

[0062] In the embodiment of the present invention, after the fire source is recognized, the visual image information and the radar response information of the target fire source are extracted from the multi-dimensional scanning information. The visual image information comes from the visible light camera and the infrared camera carried by the fire fighting cannon. The two-dimensional position of the fire source is calibrated through image processing technology, and at the same time, the high-temperature area of the fire source is determined by combining the infrared imaging data; the radar response information is generated by the lidar, and the spatial distance between the fire source and the fire fighting equipment is provided through the point cloud data. After these data are fused, the coordinate transformation method is used to calculate the three-dimensional position coordinates P(x, y, z) of the fire source, providing basic data for the subsequent range determination.

[0063] Next, the range is determined by using the pose information of the fire fighting equipment and the three-dimensional coordinates of the fire source, in combination with the range function of the fire fighting equipment. The pose information of the fire fighting equipment includes its current three-dimensional position T(x c , y c , z c ) and the attitude information. Among them, the attitude information covers the muzzle angle (horizontal angle and pitch angle) of the fire fighting cannon and the joint parameters of the boom (such as joint rotation angle and boom length, etc.), which are used to describe the complete state of the fire fighting equipment in space. This information is updated in real time by the sensor, providing dynamic input for the range model.

[0064] The fire-fighting equipment range function f(T, C, W) is a range calculation model constructed based on Bernoulli's equation, comprehensively considering multiple key influencing factors. First, the attitude information T of the fire-fighting equipment directly affects the jet direction and initial velocity of the water flow. The closer the attitude is to the direction of the fire source, the higher the jet efficiency. Second, the performance parameters C of the fire-fighting cannon include the maximum range, jet pressure, water flow type, nozzle diameter, etc. These parameters determine the dynamic characteristics of the water flow. Finally, the environmental parameters W include external factors such as wind speed, wind direction, and gravitational acceleration. These factors may cause the deviation of the water flow or the change of the range. By inputting these influencing factors into the range function, the range of the fire-fighting equipment can be calculated, and it can be judged whether the fire source is within the range.

[0065] If the fire source is within the range, a confirmation signal will be output, and the fire-fighting cannon will directly enter the fire-extinguishing mode. On the contrary, if the fire source is beyond the range, the pose adjustment module will be activated, and the target pose of the fire-fighting equipment will be planned through inverse kinematics calculation, so that the fire source enters the range. This adjustment process is dynamically updated in combination with the three-dimensional coordinates of the fire source and the range function to ensure that the fire source can always be at the center position of the range.

[0066] Based on the solution of the present invention, by fusing visual image information and radar response information, the three-dimensional coordinates of the fire source can be calculated in real time, providing high-precision fire source position data. This fusion method overcomes the defect of insufficient accuracy of a single sensor and significantly improves the reliability of fire source positioning. Based on the determination method of the fire-fighting equipment range function, the attitude, performance parameters and environmental impacts of the fire-fighting equipment are comprehensively considered, making the range calculation more accurate. Especially through the modeling of Bernoulli's equation, it can dynamically adapt to different fire field environments and ensure the accuracy of fire source determination. When the fire source is beyond the range, the target pose of the fire-fighting equipment can be quickly planned through inverse kinematics calculation, and the angles and positions of the boom and muzzle can be dynamically adjusted to make the fire source enter the range. This efficient pose adjustment strategy improves the fire-extinguishing coverage and response speed. No matter how the position of the fire source changes, the position information between the fire source and the fire-fighting equipment can be updated in real time and the range can be re-determined to ensure that the fire-fighting cannon is always in the best fire-extinguishing state. This dynamic adaptation ability is particularly important under complex fire field conditions.

[0067] Preferably, a plurality of lidars are arranged in different directions on the side of the muzzle end of the fire-fighting cannon; the determination rule of the radar response information is: based on the synchronous scanning of each lidar, the three-dimensional point cloud data of the jet area of each lidar is respectively obtained; based on the three-dimensional point cloud data of the jet area of each lidar, spatial registration is performed; based on the lidar data before and after the fire-fighting cannon sprays water of each lidar after completing spatial registration, frame difference method operation is performed to identify the jet area; based on the RANSAC algorithm, the identified jet area is fitted to obtain the fitted jet area as the radar response information.

[0068] In the embodiments of the present invention, a plurality of lidar sensors are installed on the side of the nozzle end of the fire monitor. These sensors are distributed in different directions and can perform an omnidirectional scan of the jet area, thereby providing more accurate jet area detection capabilities. The lidar sensors collect three-dimensional point cloud data of the jet area through high-frequency synchronous scanning and form multi-view jet point cloud information. Each lidar sensor independently obtains the three-dimensional point cloud data within its field of view, thus covering the entire jet area. To eliminate data inconsistencies caused by differences in the installation position and viewing angle of the sensors, a spatial registration algorithm is used to align the data of multiple lidar sensors and generate a unified three-dimensional point cloud coordinate system.

[0069] After completing the spatial registration, the registered lidar sensor data is further analyzed. Before and after the fire monitor sprays water, the lidar sensors scan the jet area and generate point cloud data respectively. By performing frame difference method operations on the data at these two moments, the dynamic changes in the jet area can be accurately identified. This method separates the jet area from the static background by detecting the changed parts in the point cloud data, thereby effectively eliminating the interference of the environmental background on jet detection.

[0070] To further improve the accuracy of jet area recognition, the RANSAC (Random Sample Consensus) algorithm is used to fit the point cloud of the extracted jet area. The RANSAC algorithm can efficiently remove noise points and fit the main trajectory curve of the jet by randomly selecting sample points and verifying their consistency. The fitting result can accurately describe the morphological characteristics and trajectory of the jet, especially the end position of the jet area. The accurate detection of this end position is crucial for the control of the jet landing point. The fitted jet area is used as the radar response information and input into the fire monitor control to provide data support for the adjustment of the jet landing point and the positioning of the fire source.

[0071] In a possible implementation, as Figure 2 , jet detection using lidar sensors is designed on both the left and right sides of the fire monitor, solving the problem of limited detection due to vision occlusion caused by factors such as jet and crosswind for a single-sided lidar sensor. Jet detection realizes the detection of the jet landing point position through two-way lidar sensor data registration, point cloud frame difference method, and RANSAC algorithm, and realizes closed-loop control of fire extinguishing based on the fire point position and the end position of the jet area.

[0072] Specifically, for the two-way lidar data acquisition and registration, the lidars on the left and right sides perform continuous scanning synchronously to collect the three-dimensional point cloud data of the jet area. The two-way lidar data is spatially registered to eliminate the inconsistency caused by the installation position difference and error. Differential operations are performed on the lidar data before the fire fighting cannon sprays water (without jet data) and after spraying water (with jet data) to find the changing areas in the point cloud, and these areas are usually the dynamic changes caused by the jet. The RANSAC algorithm of the point cloud processing technology is used to fit the point cloud in the jet area, and by analyzing and finding the end of the fitting curve, the position detection of the jet landing point is realized. According to the located fire point position and the end position of the jet area, the deviation between the two is calculated as the basis for adjusting the angle of the fire fighting cannon, and the closed-loop fire extinguishing control is realized.

[0073] Step S30: Based on the result of the determination of the position relationship, perform the pose adjustment of the fire fighting equipment until the fire fighting equipment reaches the fire extinguishing pose state.

[0074] Specifically, if the determination result of the position relationship between the fire source position and the fire fighting equipment position is that the target fire source is within the firing range of the fire fighting equipment, then directly output the result signal that the fire fighting equipment finally reaches the fire extinguishing pose state based on the current pose state of the fire fighting equipment; if the determination result of the position relationship between the fire source position and the fire fighting equipment position is that the target fire source is not within the firing range of the fire fighting equipment, then perform the pose adjustment of the fire fighting equipment based on the inverse kinematics algorithm until the fire fighting equipment reaches the fire extinguishing pose state.

[0075] Further, the performing the pose adjustment of the fire fighting equipment based on the inverse kinematics algorithm until the fire fighting equipment reaches the fire extinguishing pose state includes: determining the target pose information of the fire fighting equipment within the firing range of the fire fighting equipment based on the three-dimensional position coordinates of the target fire source and the firing range function of the fire fighting equipment; performing the inverse kinematics algorithm based on the current pose information of the fire fighting equipment and the target pose information to obtain multiple planned paths from the current pose information of the fire fighting equipment to the target pose information; performing feasibility verification on each planned path based on the kinematic constraints, filtering out the planned paths that cannot meet the kinematic constraints, and taking the planned path with the shortest time consumption among the remaining planned paths as the target path; performing the pose adjustment of the fire fighting equipment based on the target path until the fire fighting equipment reaches the fire extinguishing pose state.

[0076] In the embodiment of the present invention, two cases are processed according to the determination result of the firing range: if the fire source is within the firing range of the fire fighting equipment, directly output the fire extinguishing pose state signal based on the current pose state without further adjustment; but if the fire source exceeds the firing range, the pose of the fire fighting equipment is dynamically adjusted through the inverse kinematics algorithm until the fire source is brought into the firing range.

[0077] In the case where the pose needs to be adjusted, first, based on the three-dimensional position coordinates of the target fire source and the range function of the fire-fighting equipment, the target pose information of the fire-fighting equipment is calculated. The range function comprehensively considers the current pose information of the fire-fighting equipment (such as boom angle and muzzle direction), the performance parameters of the fire-fighting cannon (such as injection pressure and maximum range), and environmental factors (such as wind speed and wind direction), so as to dynamically determine the optimal position of the fire source within the range of the cannon. The calculation of the target pose information ensures that the fire source can be located at the center of the range, providing an accurate reference for the next adjustment.

[0078] Based on the current pose and the target pose information, the inverse kinematics algorithm is used to plan the adjustment path of the fire-fighting equipment. The inverse kinematics algorithm generates multiple possible paths from the current pose to the target pose by calculating the changes in the joint parameters of the boom of the fire-fighting equipment. During the path planning process, the kinematic constraints of each path are verified to ensure that the path is feasible in actual operation. For example, check whether the path conforms to the rotation range of the boom, the telescopic length limit, and whether it will collide with surrounding obstacles. By filtering out the paths that do not meet the constraint conditions, a set of feasible paths is finally retained.

[0079] Among the feasible paths, the target path is selected based on the principle of the shortest time consumption. This optimization strategy can not only improve the adjustment efficiency but also quickly adjust the fire-fighting equipment to the fire-extinguishing pose state in case of emergency, reducing the response time. After the target path is determined, the pose of the fire-fighting equipment is dynamically adjusted according to this path. During the adjustment process, the key parameters such as the boom angle and muzzle direction of the fire-fighting equipment are updated in real time, and the adjustment accuracy is ensured through a feedback mechanism. In particular, the actions of the boom and the muzzle are coordinated, continuously monitoring the three-dimensional position coordinates of the fire source, and combining with the real-time range determination to ensure that the fire source is always within the range.

[0080] In a possible implementation manner, according to the horizontal and vertical deflection angles Δα and Δβ of the fire-fighting cannon tracking the target fire source feedback by the control system in real time, it is input into the fire-fighting cannon controller and converted into the rotational speed control of the system for the fire-fighting cannon:

[0081]

[0082] Among them, θ is set to 2 or 3 times the control error of the fire-fighting cannon. When the deflection angle between the fire source and the fire-fighting cannon is greater than θ, the center of the fire-fighting cannon quickly approaches the center of the fire source. When the deflection angle between the fire source and the fire-fighting cannon does not exceed θ, the rotational speed S of the fire-fighting cannon is updated in real time to make the center of the fire-fighting cannon accurately approach the center of the fire source.

[0083] This pose adjustment process not only has high efficiency but also has a high degree of dynamic adaptability. When the position of the fire source changes, it can recalculate the target pose in real time and update the adjustment path, so as to adapt to the movement of the fire source and the changes in the environment. Finally, when the fire-fighting equipment reaches the target pose state and the deviation angle between the fire source position and the fire cannon is less than the set error range, an extinguishing pose state signal is output, the adjustment process is completed, and the fire-fighting equipment enters the fire-extinguishing operation mode.

[0084] Preferably, the method further includes: during the process of performing the pose adjustment of the fire-fighting equipment, scanning the target fire source in real time; when the three-dimensional position coordinates of the target fire source change, updating the target pose information of the fire-fighting equipment within the range of the fire-fighting equipment based on the new three-dimensional position of the target fire source, and updating the target path based on the updated target pose information; performing the pose adjustment of the fire-fighting equipment based on the updated target path until the fire-fighting equipment reaches the extinguishing pose state.

[0085] In the embodiment of the present invention, when the three-dimensional position coordinates of the target fire source change, the latest position of the fire source is recalculated, and combined with the current range function of the fire-fighting equipment, new target pose information is dynamically generated. The range function comprehensively considers the pose information of the fire-fighting equipment, the performance parameters of the fire cannon, and environmental factors, so as to ensure that the recalculated target pose can keep the fire source at the central position within the range.

[0086] Based on the updated target pose information, the adjustment path of the fire-fighting equipment is re-planned through the inverse kinematics algorithm. During the generation of the new path, kinematic constraint verification is performed on multiple alternative paths to filter out infeasible paths, and the optimal path (such as the path with the shortest time consumption or the highest energy efficiency) is selected for subsequent adjustment. The fire-fighting equipment adjusts the boom angle, telescopic length, and muzzle direction under the guidance of the dynamic path to achieve rapid tracking and coverage of the fire source.

[0087] During the pose adjustment process, real-time target fire source scanning is a key link. The fire-fighting equipment continuously monitors the fire source through the multi-sensors carried (including visible light cameras, infrared cameras, and lidar). These sensors work together through visual image information, temperature information, and point cloud data to ensure that the dynamic changes of the fire source can be captured in time and provide accurate input data for pose adjustment. No matter whether the fire source moves due to the expansion of the combustion range, the influence of external wind, or other reasons, it can respond quickly, update the path and target pose in real time until the fire-fighting equipment reaches the extinguishing pose state.

[0088] Figure 3 It is the system structure diagram of an intelligent fire extinguishing system provided by an embodiment of the present invention. As Figure 3As shown in the figure, an embodiment of the present invention provides an intelligent fire extinguishing system, which is realized based on the linkage control of the boom and the fire monitor of the fire fighting equipment. The system includes: a scanning unit, configured to scan for a fire source within a predetermined visible range, obtain real-time multi-dimensional scanning information, and call a corresponding fire source recognition model based on the multi-dimensional scanning information; wherein, the fire source recognition model is trained based on training sample data including multi-dimensional features of the fire source, and the multi-dimensional features of the fire source include any one or more of: fire source temperature information, fire source dynamic features, and fire source position information; a position determination unit, configured to perform a position relationship determination between the fire source position and the position of the fire fighting equipment when the fire source recognition model recognizes a fire source; a pose adjustment unit, configured to perform a pose adjustment of the fire fighting equipment based on the result of the position relationship determination until the fire fighting equipment reaches a fire extinguishing pose state.

[0089] Preferably, a sensor is provided on the side of the nozzle end of the fire monitor; the sensor includes any one or more of: a visible light camera, an infrared camera, and a lidar; a plurality of lidars are provided along different directions on the side of the nozzle end of the fire monitor.

[0090] A third aspect of the present invention provides a fire fighting equipment, which is configured with the above-mentioned intelligent fire extinguishing system.

[0091] An embodiment of the present invention further provides a computer-readable storage medium, on which instructions are stored, and when the instructions are run on a computer, the computer is caused to execute the above-mentioned intelligent fire extinguishing method.

[0092] Those skilled in the art can understand that all or part of the steps in the methods of the above embodiments can be completed by instructing relevant hardware through a program, and the program is stored in a storage medium, including several instructions for causing a single-chip microcomputer, a chip, or a processor to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as a USB flash drive, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk, or an optical disc that can store program codes.

[0093] The optional embodiments of the present invention have been described in detail above in conjunction with the accompanying drawings. However, the embodiments of the present invention are not limited to the specific details in the above embodiments. Within the scope of the technical concept of the embodiments of the present invention, various simple modifications can be made to the technical solutions of the embodiments of the present invention, and these simple modifications all fall within the protection scope of the embodiments of the present invention. In addition, it should be noted that, among the various specific technical features described in the above specific embodiments, they can be combined in any appropriate manner without contradiction. To avoid unnecessary repetition, the embodiments of the present invention will not separately describe various possible combination manners.

[0094] In addition, any combination can be made among various different embodiments of the present invention, as long as it does not violate the idea of the embodiments of the present invention, and it should also be regarded as the content disclosed by the embodiments of the present invention.

Claims

1. An intelligent fire extinguishing method, characterized in that, The method is implemented based on the linkage control of the boom and the fire monitor of the fire-fighting equipment. The method includes: Scanning for a fire source within a predetermined visible range to obtain real-time multi-dimensional scanning information, and invoking a corresponding fire-source recognition model based on the multi-dimensional scanning information. Among them, The fire-source recognition model is trained based on training sample data containing multi-dimensional fire-source features. The multi-dimensional fire-source features include any one or more of fire-source temperature information, fire-source dynamic features, and fire-source position information. When the fire-source recognition model recognizes a fire source, determine the positional relationship between the fire-source position and the position of the fire-fighting equipment. Based on the result of the positional relationship determination, perform pose adjustment of the fire-fighting equipment until the fire-fighting equipment reaches the fire-extinguishing pose state. Among them, The pose of the fire-fighting equipment includes the position information of the fire-fighting equipment, the attitude information of the boom, and the attitude information of the fire monitor.

2. The method according to claim 1, characterized in that A sensor is provided on the nozzle end side of the fire monitor. The sensor includes: Any one or more of a visible-light camera, an infrared camera, and a lidar. The multi-dimensional scanning information includes: Any one or more of video stream information, infrared image information, and radar response information.

3. The method according to claim 1, wherein The method further includes performing training of the fire-source recognition model. The training rule of the fire-source recognition model is: Collect historical image information, perform multi-dimensional fire-source feature annotation on the historical image information, and construct training samples based on the annotated images. Perform model training on a pre-constructed neural network based on the training set in the training samples. During the training process, update the network weights through backpropagation and optimization algorithms to minimize the loss function and obtain an initial model. Perform verification on the initial model based on the validation set in the training samples, and use the model that passes the verification as the fire-source recognition model.

4. The method according to claim 3, It is characterized in that; The acquisition rule of the fire-source dynamic features is: Based on the optical flow method, perform pixel displacement recognition between consecutive image frames, and obtain the dynamic features of the fire source based on the recognition result.

5. The method according to claim 3, wherein The construction rule of the training samples is: Based on a single historical image information, perform feature combination on each fire-source feature in the multi-dimensional fire-source features to obtain a set of combined features corresponding to one historical image information. Based on the combined features of all historical image information, construct corresponding training samples.

6. The method according to claim 3, wherein The pre-constructed neural network is constructed based on the YOLO network. The pre-constructed neural network includes: A combined feature input layer for inputting training samples of combined features. A convolutional feature extraction layer, including a fire-source temperature feature extraction layer and a fire-source dynamic feature extraction layer, for performing convolutional processing on the input combined features. An attention mechanism module, including an SE module or a CBAM module, for enhancing attention to features. A loss calculation module for determining the total loss based on each loss function.

7. The method according to claim 6, characterized in that, Each loss function includes: Any one or more of a classification loss function, a localization loss function, a temperature loss function, and a jump loss function. The classification loss function is used to judge whether the target area contains a flame. The localization loss function is used to optimize the coordinates and dimensions of the flame bounding box. The temperature loss function is used to predict the temperature distribution in the flame area and align it with the true value. The jump loss function is used to determine the changes in the dynamic characteristics of the flame and optimize the prediction of dynamic information.

8. The method according to claim 1, characterized in that, When the fire source recognition model recognizes a fire source, the determination of the positional relationship between the fire source position and the fire fighting equipment position is performed, including: When the fire source recognition model recognizes a fire source, the three-dimensional position coordinates of the target fire source are determined based on the combination of the visual image information and the radar response information of the target fire source in the multi-dimensional scan information; Based on the pose information of the fire fighting equipment, the three-dimensional position coordinates of the target fire source, and the fire fighting equipment range function of the fire fighting equipment, it is determined whether the target fire source is within the range of the fire fighting equipment, and the judgment result is used as the judgment result of the positional relationship between the fire source position and the fire fighting equipment position.

9. The method according to claim 8, wherein The fire fighting equipment range function is constructed based on the range calculation model of the Bernoulli equation; The influencing factors of the fire fighting equipment range function include: The attitude information of the fire fighting equipment, the performance parameter information of the fire fighting cannon, and the environmental parameter information.

10. The method according to claim 8, wherein Based on the result of the positional relationship determination, the pose adjustment of the fire fighting equipment is performed until the fire fighting equipment reaches the fire extinguishing pose state, including: If the judgment result of the positional relationship between the fire source position and the fire fighting equipment position is that the target fire source is within the range of the fire fighting equipment, the result signal that the fire fighting equipment finally reaches the fire extinguishing pose state is directly output based on the current pose state of the fire fighting equipment; If the judgment result of the positional relationship between the fire source position and the fire fighting equipment position is that the target fire source is not within the range of the fire fighting equipment, the pose adjustment of the fire fighting equipment is performed based on the inverse kinematics algorithm until the fire fighting equipment reaches the fire extinguishing pose state.

11. The method according to claim 10, wherein The pose adjustment of the fire fighting equipment is performed based on the inverse kinematics algorithm until the fire fighting equipment reaches the fire extinguishing pose state, including: Based on the three-dimensional position coordinates of the target fire source and the fire fighting equipment range function, the target pose information of the fire fighting equipment within the range of the fire fighting equipment for the target fire source is determined; Based on the current pose information of the fire fighting equipment and the target pose information, the inverse kinematics algorithm is performed to obtain multiple planned paths from the current pose information of the fire fighting equipment to the target pose information; Based on the kinematic constraints, the feasibility of each planned path is verified, the planned paths that cannot meet the kinematic constraints are filtered out, and the planned path with the shortest time consumption among the remaining planned paths is used as the target path; Based on the target path, the pose adjustment of the fire fighting equipment is performed until the fire fighting equipment reaches the fire extinguishing pose state.

12. The method according to claim 11, wherein The method further includes: During the process of performing the pose adjustment of the fire fighting equipment, the target fire source is scanned in real time; When the three-dimensional position coordinates of the target fire source change, the target pose information of the fire fighting equipment within the range of the fire fighting equipment for the target fire source is updated based on the new three-dimensional position of the target fire source, and the target path is updated based on the updated target pose information; Based on the updated target path, the pose adjustment of the fire fighting equipment is performed until the fire fighting equipment reaches the fire extinguishing pose state.

13. The method according to claim 8, wherein A plurality of lidar sensors are arranged on different sides of the nozzle end of the fire fighting cannon; The determination rule of the radar response information is: Based on the synchronous scanning of each lidar sensor, the three-dimensional point cloud data of the jet area of each lidar sensor is obtained respectively; Spatial registration is performed based on the three-dimensional point cloud data of the jet area of each lidar sensor; Perform frame difference method operations on the lidar data of each lidar before and after the fire cannon sprays water based on the completed spatial registration to identify the jet area; Fit the identified jet area based on the RANSAC algorithm to obtain the fitted jet area as the radar response information.

14. An intelligent fire extinguishing system, characterized in that, The system is implemented based on the linkage control of the boom and the fire cannon of the fire-fighting equipment. The system includes: A scanning unit for scanning for a fire source within a predetermined visible range to obtain real-time multi-dimensional scanning information and calling a corresponding fire-source recognition model based on the multi-dimensional scanning information; where The fire-source recognition model is trained based on training sample data containing multi-dimensional features of the fire source. The multi-dimensional features of the fire source include any one or more of fire-source temperature information, fire-source dynamic features, and fire-source position information; A position determination unit for determining the positional relationship between the fire-source position and the position of the fire-fighting equipment when the fire-source recognition model recognizes a fire source; A pose adjustment unit for performing pose adjustment of the fire-fighting equipment based on the result of the positional relationship determination until the fire-fighting equipment reaches the fire-extinguishing pose state; where The pose of the fire-fighting equipment includes the position information of the fire-fighting equipment, the attitude information of the boom, and the attitude information of the fire cannon.

15. The system according to claim 13, wherein A sensor is provided on the nozzle end side of the fire cannon; The sensor includes: Any one or more of a visible light camera, an infrared camera, and a lidar; A plurality of lidars are provided along different directions on the nozzle end side of the fire cannon.

16. A fire-fighting device, characterized in that, The fire-fighting equipment is configured with the intelligent fire-extinguishing system according to any one of claims 14 or 15.

17. A computer-readable storage medium, characterized in that, Instructions are stored on the computer-readable storage medium, and when running on a computer, cause the computer to execute the intelligent fire-extinguishing method according to any one of claims 1-13.

Images