Unmanned-aerial-vehicle-based intelligent fire inspection system and method

Abstract

Disclosed in the present invention are an unmanned-aerial-vehicle-based intelligent fire inspection system and method. The system comprises a design module, a capture module, a transmission module, an identification module, and an alarm module, wherein the design module is configured to design an inspection path of an unmanned aerial vehicle; the capture module is configured to perform inspection according to the inspection path, capture inspection scene images and videos, and temporarily store image and video data; the transmission module is configured to transmit the image and video data to a command center; the identification module is configured to identify a fire safety status in the image and video data and lock a target area; and the alarm module is configured to: when a fire occurs, issue alarm information and trigger a fire early-warning handling mechanism. The present invention can cooperate with an automatic takeoff and landing platform to realize 24 / 7 standby, can quickly respond to fire incidents, can reduce the risks for on-site personnel, can improve the overall efficiency of fire rescues, and can provide a strong support for safeguarding people's lives and property.

Description

A drone-based intelligent fire inspection system and method Technical Field

[0001] This invention relates to the field of fire inspection, specifically to an unmanned aerial vehicle (UAV) intelligent fire inspection system and method. Background Technology

[0002] In modern society, sudden disasters such as fires pose a serious threat to people's lives and property. Traditional fire inspection methods mainly rely on manual patrols, which have many limitations. On the one hand, manual patrols are inefficient and cannot provide comprehensive and timely coverage of large areas; on the other hand, complex terrain or dangerous environments pose significant safety hazards to patrol personnel and make it difficult to accurately and quickly identify fire hazards. In addition, traditional fire protection systems have a relatively slow emergency response speed after a fire breaks out, making it impossible to obtain detailed information from the scene in a timely manner, thus affecting the efficiency of rescue decision-making and execution.

[0003] The rapid development of drone technology, artificial intelligence technology, and communication technology has provided new ideas and methods for solving the above problems. Drones have advantages such as maneuverability, rapid access to designated areas, and the ability to carry multiple sensors, enabling efficient inspection of target areas. Meanwhile, artificial intelligence technology has made breakthroughs in image recognition, enabling accurate and rapid identification of fire features in images. These technologies, when applied to fire inspection systems, provide strong support for early fire warnings and emergency response. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes an intelligent fire inspection system and method using unmanned aerial vehicles (UAVs), aiming to build an efficient, intelligent, and reliable fire inspection solution to improve the efficiency and effectiveness of fire prevention and emergency rescue, and safeguard people's lives and property.

[0005] To achieve the above objectives, the present invention provides an intelligent fire inspection system for unmanned aerial vehicles (UAVs), comprising: a design module, a shooting module, a transmission module, an identification module, and an alarm module;

[0006] The design module is used to plan and design the inspection path of the UAV;

[0007] The camera module is used to perform inspections based on the designed inspection path, and to capture images and videos of the inspection scene, and temporarily store the image and video data.

[0008] The transmission module is used to transmit the image and video data to the command center;

[0009] The identification module is used to identify the fire status in the image and video data and lock the target area;

[0010] The alarm module is used to issue an alarm and trigger the fire early warning and response mechanism when a fire occurs.

[0011] Preferably, the design module includes: a data acquisition unit, a scene construction unit, and a path planning unit;

[0012] The data acquisition unit is used to collect site information of the area to be inspected.

[0013] The scene construction unit is used to construct a simulated scene of the area to be inspected based on the site information;

[0014] The path planning unit is used to plan inspection paths based on the simulated scenario.

[0015] Preferably, the A-Star algorithm is used for path planning, and its expression is as follows: f(n) = g(n) + h(n)

[0016] Where g(n) represents the actual cost from the starting point to the current node n; h(n) represents the heuristically estimated cost from the current node n to the target node; and f(n) represents the total evaluation cost, used to determine the priority of nodes.

[0017] Preferably, the scene construction unit uses a pre-trained convolutional neural network model to construct a fire recognition model; key features of the fire image are extracted through multi-layer convolution and pooling operations; during the training process of the convolutional neural network, the number of network layers, convolution kernel size, and activation function are set according to the characteristics of the fire recognition task, and the model is optimized using the cross-entropy loss function and stochastic gradient descent optimization algorithm.

[0018] Preferably, the A-star algorithm is used for path planning. The structure of the fire identification model includes: the first convolutional layer uses 32 convolutional kernels of size 3x3; the second and third convolutional layers use 64 convolutional kernels of size 3x3; the pooling layer has a window size of 3x3; and the activation function is the ReLU function.

[0019] This invention also provides an intelligent fire inspection method using unmanned aerial vehicles (UAVs), the method being used to implement the above-mentioned system, the steps of which include:

[0020] Planning and designing inspection paths for drones;

[0021] Conduct inspections according to the designed inspection route, and take inspection photos and videos, and temporarily store the image and video data;

[0022] The image and video data are transmitted to the command center;

[0023] Identify the fire status in the video data and lock the target area.

[0024] Preferably, the A-Star algorithm is used for path planning, and its expression is as follows: f(n) = g(n) + h(n)

[0025] Where g(n) represents the actual cost from the starting point to the current node n; h(n) represents the heuristically estimated cost from the current node n to the target node; and f(n) represents the total evaluation cost, used to determine the priority of nodes.

[0026] Preferably, a pre-trained convolutional neural network model is used to construct a fire identification model; key features of fire images are extracted through multi-layer convolution and pooling operations; during the training process of the convolutional neural network, the number of network layers, convolutional kernel size, and activation function are set according to the characteristics of the fire identification task, and the model is optimized using the cross-entropy loss function and stochastic gradient descent optimization algorithm.

[0027] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0028] This invention can be used in conjunction with an automated take-off and landing platform to achieve 24 / 7 standby, rapid response to fire incidents, reduce risks to on-site personnel, improve the overall efficiency of fire rescue, and provide strong support for protecting people's lives and property. Attached Figure Description

[0029] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 is a schematic diagram of the system structure according to an embodiment of the present invention. Detailed Implementation

[0031] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0032] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0033] Example 1

[0034] Figure 1 shows the system structure flowchart of this embodiment, which includes: a design module, a shooting module, a transmission module, an identification module, and an alarm module; the design module is used to plan and design the inspection path of the UAV; the shooting module is used to perform inspections based on the designed inspection path, and to shoot images and videos of the inspection scene, and temporarily store the image and video data; the transmission module is used to transmit the image and video data to the command center; the identification module is used to identify the fire situation in the image and video data and lock the target area; the alarm module is used to issue an alarm and trigger the fire early warning and response mechanism when a fire occurs.

[0035] The following will describe in detail, with reference to this embodiment, how the present invention solves the technical problems in practical work.

[0036] First, the data acquisition unit in the design module collects site information for the area to be inspected, including surrounding environmental information, terrain, and geographic information. The surrounding environmental information includes vegetation (such as trees and grass), shading (specifically, areas with dim or absent light due to objects or terrain obstruction), and obstacles (such as buildings, walls, and poles). These factors significantly influence the UAV's flight path and obstacle avoidance, aiding the planning module in path planning and flight safety assessment.

[0037] Topographic and geographical information includes: elevation, flatness, and undulation. Considering the drone's flight altitude and speed during inspections, and understanding the slope and undulation of the site, can help avoid unnecessary shaking or collision risks during takeoff, landing, or flight.

[0038] Then, based on the collected site information, the scene construction unit uses BIM to construct a simulated scene of the area to be inspected. By importing the above parameters into the BIM software, a 3D model is created to obtain the simulated scene.

[0039] The path planning unit then uses the A-star algorithm to plan the path based on the constructed scenario, and its expression is as follows: f(n) = g(n) + h(n)

[0040] Where g(n) represents the actual cost from the starting point to the current node n; h(n) represents the heuristically estimated cost from the current node n to the target node; and f(n) represents the total evaluation cost, used to determine the priority of nodes.

[0041] After the path planning is completed, the drone will carry out inspections according to the planned path. In this embodiment, the shooting module is a high-definition camera mounted on the drone, which is used to capture high-definition images or videos of the inspection area.

[0042] After filming is completed, the images and videos are transmitted to the command center using a transmission module that employs 5G technology, allowing the command center to obtain real-time images and video feeds from the drone.

[0043] Next, a fire identification model is constructed using the identification module to identify fires in the acquired images, video streams, or image data. Specifically, this embodiment uses a pre-trained convolutional neural network (CNN) model to construct the fire identification model. In the CNN, key features of the fire image are extracted through multi-layer convolution and pooling operations. During the CNN training process, appropriate hyperparameters such as the number of network layers, kernel size, and activation function are set according to the characteristics of the fire identification task. The model is then optimized using the cross-entropy loss function and stochastic gradient descent optimization algorithm. The trained CNN model is then evaluated using a test set. By calculating indicators such as recognition accuracy and false recognition rate, the model's generalization ability and robustness under different lighting conditions are assessed. If the model performance does not meet expectations, the model is improved by adjusting the network structure, optimizing the algorithm, and increasing training data. This training and evaluation process is repeated until a fire identification model that meets the robustness requirements is obtained.

[0044] Specifically, the model structure in this embodiment is as follows:

[0045] The first convolutional layer uses 32 3x3 kernels to extract low-level features such as edges and textures from the fire image. The second and third convolutional layers use 64 3x3 kernels to extract higher-level features, such as smoke and flames. The pooling layer has a 3x3 window size, which reduces the dimensionality of the feature map, decreases computation, and improves the model's robustness. It also captures a larger receptive field. The ReLU activation function is used.

[0046] By calculating metrics such as recognition accuracy and false recognition rate, the generalization ability and robustness of the model under different lighting conditions are assessed. For example, if the model achieves a recognition accuracy of over 95% and a false recognition rate of less than 5% on the test set, the model is considered to perform well. If the model performance does not meet expectations, model improvement is necessary.

[0047] After the identification module completes the identification, the alarm module activates the corresponding fire early warning and response mechanism based on the severity and damage of the fire, and notifies firefighters to arrange and dispatch tasks.

[0048] Furthermore, the firefighting drone in this embodiment can be deployed in conjunction with an automated take-off and landing platform. This platform supports 24 / 7 drone standby and can be deployed within 5 minutes of receiving an order, significantly shortening response time. The drone can penetrate deep into narrow and dangerous areas to reconnoiter the environment, reducing risks to on-site personnel; it can transmit real-time image information to provide reference for rescue decisions; and it can carry various rescue equipment such as fire water tanks and transport cabins to carry out operations.

[0049] Example 2

[0050] This embodiment also provides an intelligent fire inspection method using unmanned aerial vehicles (UAVs). The method is used to implement the above-mentioned system and includes the following steps: designing the inspection path of the UAV; conducting inspections according to the designed inspection path and capturing images and videos of the inspection scene, and temporarily storing the image and video data; transmitting the image and video data to the command center; identifying the fire status in the image and video data, and locking the target area.

[0051] The A-Star algorithm is used for path planning, and its expression is as follows: f(n) = g(n) + h(n)

[0052] Where g(n) represents the actual cost from the starting point to the current node n; h(n) represents the heuristically estimated cost from the current node n to the target node; and f(n) represents the total evaluation cost, used to determine the priority of nodes.

[0053] A fire identification model is constructed using a pre-trained convolutional neural network model. Key features of fire images are extracted through multi-layer convolution and pooling operations. During the training process of the convolutional neural network, the number of network layers, convolutional kernel size, and activation function are set according to the characteristics of the fire identification task. The model is then optimized using the cross-entropy loss function and the stochastic gradient descent optimization algorithm.

[0054] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A drone-based intelligent fire inspection system, characterized in that, include: The system includes a design module, a shooting module, a transmission module, a recognition module, and an alarm module. The design module is used to design the inspection path for the UAV; The camera module is used to perform inspections based on the designed inspection path, capture scenes, and generate and temporarily store image and video data. The transmission module is used to transmit the image and video data to the command center; The identification module is used to identify the fire status in the image and video data and lock the target area; The alarm module is used to issue an alarm and trigger the fire early warning and response mechanism when a fire occurs.

2. The unmanned aerial vehicle (UAV) intelligent fire inspection system according to claim 1, characterized in that, The design module includes: a data acquisition unit, a scene construction unit, and a path planning unit; The data acquisition unit is used to collect site information of the area to be inspected. The scenario construction unit is used to construct a simulated scenario of the area to be inspected based on the inspection site information; The path planning unit is used to plan inspection paths based on the simulated scenario.

3. The unmanned aerial vehicle (UAV) intelligent fire inspection system according to claim 2, characterized in that, The A-Star algorithm is used for path planning, and its expression is as follows: f(n) = g(n) + h(n) Where g(n) represents the actual cost from the starting point to the current node n; h(n) represents the heuristically estimated cost from the current node n to the target node; and f(n) represents the total evaluation cost, used to determine the priority of nodes.

4. The unmanned aerial vehicle (UAV) intelligent fire inspection system according to claim 2, characterized in that, The scene construction unit uses a pre-trained convolutional neural network model to build a fire recognition model; it extracts key features of fire images through multi-layer convolution and pooling operations; during the training process of the convolutional neural network, it sets the number of network layers, convolutional kernel size, and activation function according to the characteristics of the fire recognition task, and uses the cross-entropy loss function and stochastic gradient descent optimization algorithm to optimize the model.

5. The unmanned aerial vehicle (UAV) intelligent fire inspection system according to claim 4, characterized in that, The A-star algorithm is used for path planning. The structure of the fire identification model includes: the first convolutional layer uses 32 convolutional kernels of size 3x3; the second and third convolutional layers use 64 convolutional kernels of size 3x3; the pooling layer has a window size of 3x3; and the activation function is the ReLU function.

6. A method for intelligent fire inspection using unmanned aerial vehicles (UAVs), the method being used to implement the system described in any one of claims 1-5, characterized in that the steps include... include: Design inspection paths for drones; Conduct inspections according to the designed inspection route, and take pictures and videos of the inspection scene, and temporarily store the image and video data. The image and video data are transmitted to the command center; Identify the fire situation in the image and video data and locate the target area.

7. The intelligent fire inspection method using unmanned aerial vehicles (UAVs) according to claim 6, characterized in that, The A-Star algorithm is used for path planning, and its expression is as follows: f(n) = g(n) + h(n) Where g(n) represents the actual cost from the starting point to the current node n; h(n) represents the heuristically estimated cost from the current node n to the target node; and f(n) represents the total evaluation cost, used to determine the priority of nodes.

8. The intelligent fire inspection method using unmanned aerial vehicles (UAVs) according to claim 6, characterized in that, A fire identification model is constructed using a pre-trained convolutional neural network model. Key features of fire images are extracted through multi-layer convolution and pooling operations. During the training process of the convolutional neural network, the number of network layers, convolutional kernel size, and activation function are set according to the characteristics of the fire identification task. The model is then optimized using the cross-entropy loss function and the stochastic gradient descent optimization algorithm.

Images