Non-coal mine intelligent inspection method and intelligent inspection unmanned vehicle

Through intelligent inspection of unmanned vehicles combined with environmental data and image recognition technology, the inspection route is adjusted, and the problem of limited efficiency and safety in complex mining environments is solved, and efficient and safe mine inspection is achieved.

CN118938941AInactive Publication Date: 2024-11-12WUHAN DIDA HUARUI GEOSCIENCE TECH CO LTD
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
CN202411418668.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-12
Publication Date
2024-11-12
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Traditional mine inspection methods rely on manual or fixed route inspection vehicles, making it difficult to adapt to complex and dangerous mining environments, resulting in limited inspection efficiency and safety.

Method used

Intelligent inspection of unmanned vehicles is adopted to obtain mining maps, collect environmental and image data in real time, adjust the inspection route, and combine technologies such as permeability detection and gas sensors to judge and adjust the inspection route in real time to ensure safety and efficiency.

Benefits of technology

Efficient and safe inspections in complex and dangerous mining environments have been achieved, inspection efficiency and safety have been improved, and the risks of manual intervention have been reduced.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118938941A_ABST
    Figure CN118938941A_ABST
Patent Text Reader

Abstract

The invention discloses a non-coal mine intelligent inspection method and an intelligent inspection unmanned vehicle, and the method comprises the steps: obtaining a mine region map of a mine, and marking a plurality of dangerous regions in the mine region map in advance; determining a plurality of target inspection points according to the plurality of dangerous areas, and generating an initial inspection route based on the target inspection points; controlling the intelligent tour-inspection unmanned vehicle to move towards the target tour-inspection point, collecting environment data and image data of the surrounding environment in real time, and adjusting a tour-inspection route according to the environment data and the image data; after any inspection point is reached, whether the current inspection point meets a detection condition or not is judged by combining the environment data and the image data; if the detection condition is met, performing water permeation detection, and analyzing to obtain a water permeation hidden danger result of the current inspection point; and judging whether the current inspection point has a danger symptom according to the water permeation hidden danger result and the environmental data and determining the danger level.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The embodiments of the present application relate to the field of intelligent inspection, and in particular to a non-coal mine intelligent inspection method and an intelligent inspection unmanned vehicle. Background Art

[0002] As a high-risk production environment, mine safety management has always been an important part of mine operations. Traditional mine safety management mainly relies on manual inspection and monitoring. Although this method can discover and deal with safety hazards to a certain extent, it has many limitations and shortcomings. The mine environment usually includes narrow passages, unstable geological structures, and potential dangerous factors such as toxic gases and accumulated water, which pose severe challenges to manual inspections. Traditional manual inspections not only require inspectors to work for long periods of time in harsh environments, but also require them to have high professional skills and rich experience to identify and deal with various potential dangers. With the increase in mining depth and the expansion of production scale, the mine environment has become more and more complex and dangerous, and the traditional manual inspection method has been unable to meet the needs of modern mine safety management.

[0003] In the existing technology, patrol vehicles are usually used to replace manual inspections, but the navigation system of conventional patrol vehicles often relies on preset fixed routes and lacks flexibility and adaptability. When the internal structure of the mine changes or new obstacles appear, the patrol vehicle cannot adjust the route in time, which may cause deviation from the designated inspection point and failure to complete the scheduled inspection task, which will eventually seriously affect the efficiency and safety of the inspection work. Summary of the invention

[0004] The embodiments of the present application provide a non-coal mine intelligent inspection method and an intelligent inspection unmanned vehicle, which are used to solve the problem of low inspection efficiency of unmanned inspection vehicles in the prior art.

[0005] To achieve the above objectives, in a first aspect, the present application provides a non-coal mine intelligent inspection method, the method comprising: Obtaining a mine area map of a target mine area, wherein the mine area map is pre-marked with a plurality of dangerous areas; Determine a plurality of target inspection points according to the plurality of dangerous areas, and generate an initial inspection route based on all the target inspection points; Controlling the intelligent inspection unmanned vehicle to move toward the target inspection point along the initial inspection route; during the movement of the intelligent inspection unmanned vehicle toward the target inspection point, collecting environmental data and image data of the surrounding environment in real time through the intelligent inspection unmanned vehicle, and adjusting the initial inspection route to the target inspection route in combination with the environmental data and the image data; Controlling the intelligent inspection unmanned vehicle to move toward the target inspection point according to the target inspection route; After arriving at any of the target inspection points, judging whether the current target inspection point meets the detection conditions by combining the environmental data and the image data; If the current target inspection point meets the detection conditions, the intelligent inspection unmanned vehicle performs water seepage detection on the current target inspection point to obtain a preliminary water seepage detection result, and combines the environmental data and the preliminary water seepage detection result to obtain a water seepage hazard result of the current target inspection point; Determine whether the target inspection point currently has any danger signs according to the water seepage hazard result and the environmental data; The environmental data and the water penetration detection result are combined to analyze and determine the current danger level of the target inspection point.

[0006] Optionally, the intelligent inspection unmanned vehicle includes a lifting device disposed on the top of the intelligent inspection unmanned vehicle, and also includes a water penetration detection module disposed on the top of the lifting device, wherein the intelligent inspection unmanned vehicle performs water penetration detection on the current target inspection point to obtain a preliminary water penetration detection result, and combines the environmental data and the preliminary water penetration detection result to analyze and obtain the water penetration hazard result of the current target inspection point, including the following steps: Control the lifting device to lift the water permeability detection module to the top of the mine cave at the current target inspection point, and use the transient electromagnetic instrument coil in the water permeability detection module to detect the rock mass at the top of the mine cave to obtain preliminary detection data; Analyze the preliminary detection data to obtain the resistivity value of the rock mass in the mine at the current target inspection point; Analyze the resistivity value to obtain the preliminary water penetration detection result of the current target inspection point; The environmental data and the preliminary water seepage detection result are analyzed in combination to obtain the water seepage hazard result of the current target inspection point.

[0007] Optionally, the environmental data includes environmental humidity data, and the combining the environmental data and the preliminary water permeability detection result to obtain the water permeability hazard result of the current target inspection point includes the following steps: Construct a primary mine hidden danger identification model based on RBF neural network model; Obtain historical environmental humidity data and historical water penetration detection results; Inputting the historical environmental humidity data and the historical water permeability detection results into the primary mine cave hidden danger identification model, training the primary mine cave hidden danger identification model, and obtaining a mine cave hidden danger identification model; Inputting the environmental humidity data and the preliminary water permeability detection result into the trained mine hazard identification model, mapping the environmental humidity data and the preliminary water permeability detection result into a latent space to obtain latent space information; Based on the latent space information and through the mine hazard identification model, the water seepage hazard result of the target inspection point is output.

[0008] Optionally, the environmental data further includes harmful gas concentration data, and judging whether the current target inspection point has any dangerous signs according to the water seepage hazard result and the environmental data includes the following steps: If the water seepage hidden danger result is that there is a water seepage hidden danger, it is determined that the current target inspection point has a dangerous sign; If the water seepage hidden danger result is that there is no water seepage hidden danger, judging whether there is any danger sign at the current target inspection point according to the harmful gas concentration data; If the harmful gas concentration data is less than the preset harmful gas concentration threshold, it is determined that there is no danger sign at the current target inspection point; If the harmful gas concentration data is greater than or equal to a preset harmful gas concentration threshold, it is determined that a dangerous sign appears at the current target inspection point.

[0009] Optionally, controlling the intelligent inspection unmanned vehicle to move toward the target inspection point along the initial inspection route, collecting environmental data and image data in real time by the intelligent inspection unmanned vehicle during the movement of the intelligent inspection unmanned vehicle toward the target inspection point, and adjusting the initial inspection route to the target inspection route in combination with the environmental data and the image data comprises the following steps: Controlling the intelligent inspection unmanned vehicle to move toward the target inspection point according to the initial inspection route; In the process of the intelligent inspection unmanned vehicle moving toward the target inspection point, the intelligent inspection unmanned vehicle collects environmental data and image data in real time, wherein the environmental data includes light intensity data; If the light intensity data is greater than a preset light intensity threshold, the image recognition technology is used to identify the mine cave terrain features in the image data, and the initial inspection route is adjusted based on the mine cave terrain features; If the light intensity data is less than or equal to a preset light intensity threshold, the initial inspection route is adjusted to a target inspection route.

[0010] Optionally, the identifying the mine cave terrain features in the image data by using image recognition technology, and adjusting the initial inspection route based on the mine cave terrain features comprises the following steps: Preprocessing the image data into a grayscale image; A threshold segmentation algorithm is used to separate the target image region including the complete mine tunnel entrance from the image background in the grayscale image; Locating the target image area by using a local binary algorithm, and calculating the area size of the target image area; Obtaining the camera parameters of the intelligent inspection unmanned vehicle when collecting image data, and using the on-board laser ranging device of the intelligent inspection unmanned vehicle to measure the actual distance between the intelligent inspection unmanned vehicle and the mine entrance; The actual size of the mine entrance is calculated by combining the area size, the camera parameters and the actual distance; If the actual size of the hole does not match the vehicle size of the intelligent inspection unmanned vehicle, the initial inspection route is adjusted to a target inspection route; If the actual size of the hole matches the vehicle size of the intelligent inspection unmanned vehicle, image enhancement processing is performed on the target image area; Combining the threshold segmentation algorithm and the local binary algorithm to identify and analyze the target image area after image enhancement processing, and judging whether there are obstacles inside the target mine cave to which the mine cave entrance belongs that hinder the advancement of the intelligent inspection unmanned vehicle; If there is no obstacle inside the target mine cave to which the mine cave entrance belongs that hinders the advance of the intelligent inspection unmanned vehicle, the intelligent inspection unmanned vehicle is controlled to travel along the initial inspection route; If there are obstacles inside the target mine cave to which the mine cave entrance belongs that hinder the advancement of the intelligent inspection unmanned vehicle, the initial inspection route is adjusted to the target inspection route.

[0011] Optionally, the environmental data further includes corrosive gas concentration data, and the combining of the environmental data and the image data to determine whether the current target inspection point meets the detection condition includes the following steps: Extract dust images from image data using image recognition technology; Calculating the saturation and brightness of the dust image to obtain the light transmittance of the dust image; Analyzing and processing the light transmittance of the dust image to obtain the dust concentration characteristics; The dust concentration characteristics, the environmental humidity data and the corrosive gas concentration data are combined, and a comprehensive analysis is performed through a kernel principal component analysis method to determine whether the current target inspection point meets the detection conditions.

[0012] Optionally, combining the dust concentration characteristics, the humidity data and the corrosive gas concentration data, and comprehensively analyzing whether the current target inspection point meets the detection conditions through a kernel principal component analysis method includes the following steps: Mapping the dust concentration characteristics, the humidity data and the corrosive gas concentration data to the same high-dimensional feature space; Extracting fusion data features from the high-dimensional feature space using principal component analysis; Construct a mine dangerous environment identification model based on BP neural network model; Inputting the fused data features into a preset mine dangerous environment identification model, and outputting a dangerous environment identification result through the mine dangerous environment identification model; Determining whether the current environment of the intelligent inspection unmanned vehicle has a corrosive effect on the electronic equipment of the intelligent inspection unmanned vehicle according to the dangerous environment identification result; If the environment in which the intelligent inspection unmanned vehicle is located does not have a corrosive effect on the electronic equipment of the intelligent inspection unmanned vehicle, it is determined that the current target inspection point meets the detection conditions; If the environment in which the intelligent inspection unmanned vehicle is located has a corrosive effect on the electronic equipment of the intelligent inspection unmanned vehicle, it is determined that the current target inspection point does not meet the detection conditions.

[0013] Optionally, the extracting fusion data features from the high-dimensional feature space using principal component analysis method comprises the following steps: Performing an orthogonal transformation on the data in the high-dimensional feature space to obtain principal components; The principal components are screened according to a preset condition threshold to obtain the fused data features.

[0014] In a second aspect, the present application provides an intelligent inspection unmanned vehicle device, comprising: A base, with a moving device for movement arranged at the bottom; A lifting device, arranged on the top of the base; A water permeability detection module is arranged on the top of the lifting device and is used to measure the resistivity value of the rock mass in the mine; Gas sensors are arranged on both sides of the base to detect the concentration data of harmful gases in the surrounding environment; Humidity sensors are arranged on both sides of the base and are used to detect ambient humidity data of the surrounding environment; Light-sensitive sensors are arranged on both sides of the base and are used to detect light intensity data of the surrounding environment; A high-definition camera is arranged in front of the intelligent inspection unmanned vehicle and is used to collect image data of the surrounding environment.

[0015] Through the above technical solution, the present application obtains the mine area map of the target mine area in advance, marks the dangerous area on the map, determines multiple target inspection points according to the marked dangerous area, and generates an initial inspection route based on these target inspection points. The intelligent inspection unmanned vehicle is controlled to move to the inspection point according to the initial inspection route, and the environmental data and image data on the current inspection route are collected in real time during the movement. The collected environmental data and image data are analyzed to determine whether the current inspection route is suitable for continued driving. If not, the route is adjusted in combination with the collected data, and then the intelligent inspection unmanned vehicle is controlled to continue to move to the inspection point according to the readjusted route. The above steps are repeated during the movement until the intelligent inspection unmanned vehicle reaches the target inspection point. After arriving at the inspection point, the environmental data and image data of the inspection point are collected to determine whether the current inspection point is suitable for detection. If not, go to the next inspection point. If it is suitable for detection, the intelligent inspection unmanned vehicle is controlled to perform water permeability detection on the inspection point to obtain preliminary water permeability detection results. Combine the preliminary water seepage detection results and environmental data to determine whether there are any danger signs at the current inspection point. If there are any danger signs, output the danger level of the current inspection point.

[0016] Other features and advantages of the embodiments of the present application will be described in detail in the subsequent specific implementation section. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A schematic diagram of a process flow of a non-coal mine intelligent inspection method provided in an embodiment of the present application; Figure 2 A system framework schematic diagram of an intelligent patrol unmanned vehicle and unmanned vehicle associated modules is provided for an embodiment of the present application. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the specific implementation methods described herein are only used to illustrate and explain the embodiments of the present application, and are not used to limit the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0019] It should be noted that if the embodiments of the present application involve directional indications (such as up, down, left, right, front, back...), such directional indications are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0020] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of the present application, the descriptions of "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or suggesting their relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by this application.

[0021] Figure 1 The following schematically shows a flow chart of a non-coal mine intelligent inspection method according to an embodiment of the present application. Figure 1 As shown, the embodiment of the present application provides a method for intelligent inspection of non-coal mines, the method comprising the following steps: S101. Obtain a mine area map through a mine management system of a target mine area, where a plurality of dangerous areas are pre-marked on the mine area map.

[0022] In the present application example, the target mining area is mainly non-coal mines, and the dangerous area refers to the area where there are hidden dangers of water seepage, poisoning, etc., which may cause safety accidents during mining work. The marking of the dangerous area is manually marked based on previous manual detection experience. S102: Determine multiple target inspection points according to multiple dangerous areas, and generate an initial inspection route based on all target inspection points.

[0023] In this embodiment, the center point of each dangerous area is used as the target inspection point, and then the initial inspection route is generated based on these target inspection points in combination with the shortest path algorithm, which includes the Dijkstra algorithm, the Bellman-Ford algorithm, the Floyd algorithm, etc. For example, the shortest inspection route obtained by using the Floyd algorithm includes the following steps: Step 1.1: Randomly select three adjacent inspection points from all inspection points and mark them as k, v, and w respectively; Step 1.2: k is the transit point, v is the starting point, w is the end point, matrix D[v][w] represents the shortest path between v and w, matrix D[v][k] represents the shortest path between v and k, and matrix D[k][w] represents the shortest path between k and w. Loop to compare the minimum values ​​of D[v][w] and D[v][k]+D[k][w]. If D[v][k]+D[k][w] is the smaller value, then overwrite D[v][k]+D[k][w] and save it in D[v][w]. Step 1.3: Repeat steps 1.1 and 2.1 until all inspection points are traversed and the shortest inspection route between all inspection points is finally obtained.

[0024] Floyd's algorithm can use the idea of ​​dynamic programming to find the shortest distance between all inspection points, allowing the intelligent inspection unmanned vehicle to reach the inspection point in the shortest time, which not only improves the working efficiency of the intelligent inspection unmanned vehicle, but also reduces the energy consumption of the intelligent inspection unmanned vehicle.

[0025] S103, control the intelligent inspection unmanned vehicle to move toward the target inspection point along the initial inspection route. During the movement of the intelligent inspection unmanned vehicle toward the target inspection point, the intelligent inspection unmanned vehicle collects environmental data and image data of the surrounding environment in real time, and adjusts the initial inspection route to the target inspection route in combination with the environmental data and image data.

[0026] In this embodiment, after receiving the inspection route and inspection instructions, the intelligent inspection unmanned vehicle moves to the target inspection point according to the received inspection route. During the driving process, the intelligent inspection unmanned vehicle will use sensors and high-definition cameras to collect real-time environmental data and image data on the driving route. The environmental data includes air humidity, air composition, etc.

[0027] The intelligent inspection unmanned vehicle sends the collected environmental data and image data to the edge computing unit, which analyzes the collected data to detect whether the air humidity exceeds the standard, whether there are corrosive gases in the air, whether the light intensity is sufficient, and whether there are obstacles on the route. After the analysis is completed, the edge computing unit sends the analysis results to the domain controller. If the current inspection route has problems such as excessive air humidity, corrosive gases, low light intensity, or obstacles on the route, the domain controller sends a route adjustment instruction to the intelligent inspection unmanned vehicle to control the intelligent inspection unmanned vehicle to re-plan the route.

[0028] Specifically, if the air humidity exceeds the standard, it will affect the performance of the intelligent inspection unmanned vehicle. Similarly, the presence of corrosive gases in the air will also affect the equipment of the intelligent inspection unmanned vehicle. If the excessive air humidity or the presence of corrosive gases in the air is not detected in time, changing the inspection route may affect the subsequent detection work, resulting in inaccurate subsequent detection results and serious consequences. The purpose of detecting light intensity is to make the images collected during image acquisition clearer. If the light intensity is too low, it will affect the image acquisition process and interfere with the intelligent inspection unmanned vehicle's accurate judgment of terrain features. Therefore, before detecting terrain features, the light intensity must be detected first. When the light intensity is sufficient, the terrain features of the current target inspection point are detected. The terrain features mainly include whether the current mine cave size can accommodate the smooth passage of the intelligent inspection unmanned vehicle, and whether there are obstacles on the current route.

[0029] S104: Control the intelligent inspection unmanned vehicle to move toward the target inspection point along the target inspection route.

[0030] In this embodiment, new driving route information and driving instructions are sent to the intelligent inspection unmanned vehicle through the domain controller, and the intelligent inspection unmanned vehicle is controlled to move to the target inspection point according to the new target inspection route.

[0031] S105: After arriving at any target inspection point, determine whether the current target inspection point meets the detection conditions by combining the environmental data and the image data.

[0032] In this embodiment, after the intelligent inspection unmanned vehicle arrives at the target inspection point, it will again use sensors and high-definition cameras to collect environmental data and image data of the current target inspection point.

[0033] Specifically, the intelligent inspection unmanned vehicle will send the collected environmental data and image data to the edge computing unit, which will analyze the environmental data and image data collected by the intelligent inspection unmanned vehicle to determine whether the current inspection point is suitable for inspection.

[0034] If the current inspection point has any of the following conditions, such as excessive air humidity or the presence of corrosive gases, it means that the inspection point is not suitable for inspection. The reason is that excessive air humidity and corrosive gases will affect the performance of the intelligent inspection unmanned vehicle and may even cause equipment damage and property losses.

[0035] S106. If the current target inspection point meets the detection conditions, the current target inspection point is subjected to water seepage detection by an intelligent inspection unmanned vehicle to obtain a preliminary water seepage detection result, and the water seepage hazard result of the current target inspection point is obtained by combining the environmental data and the preliminary water seepage detection result.

[0036] In this embodiment, refer to Figure 2After the edge computing unit is used to determine that the current target inspection point meets the detection conditions, the edge computing unit is used to send the judgment result to the domain controller. After receiving the judgment result through the domain controller, the domain controller is used to send the water seepage detection start instruction to the intelligent inspection unmanned vehicle, and the intelligent inspection unmanned vehicle is controlled to start water seepage detection. After the detection is completed, the intelligent inspection unmanned vehicle sends environmental data and preliminary water seepage detection results to the edge computing unit, and then the edge computing unit is used to analyze the environmental data and preliminary water seepage detection results to obtain the water seepage hidden danger results of the current target inspection point. After the analysis is completed, the water seepage hidden danger results are sent to the domain controller through the edge computing unit, and then sent to the inspection task analysis platform of the mine dispatching center through the domain controller, and displayed by the inspection task analysis platform of the mine dispatching center. At the same time, the domain controller sends the elevator recovery instruction to the intelligent inspection unmanned vehicle, and then controls the lifting device of the intelligent inspection unmanned vehicle to recover the elevator. After the elevator is successfully recovered, the intelligent inspection unmanned vehicle is controlled to send a recovery success instruction to the domain controller.

[0037] S107. Determine whether there are any danger signs at the current target inspection point based on the water seepage hazard results and environmental data.

[0038] In this embodiment, if the edge computing unit analyzes the current target inspection point based on the water seepage hazard results and environmental data and finds that a water seepage hazard or a poisoning hazard exists, it is determined that the current target inspection point has dangerous signs.

[0039] Specifically, water seepage hazards are mainly judged based on data such as air humidity and internal resistivity of the mine rock mass.

[0040] The risk of poisoning is determined based on the concentration of toxic gases in the mine air.

[0041] S108. Analyze and determine the danger level of the current target inspection point in combination with environmental data and water seepage detection results.

[0042] In this embodiment, after the edge computing unit sends the inspection result of the target inspection point showing dangerous signs to the domain controller, the domain controller classifies the dangerous signs and generates a danger level alarm signal.

[0043] Specifically, the hazard level is divided according to the severity of the water seepage hazard. First, determine whether the minimum value of the water seepage detection data is greater than 15. m, if greater than 15 ·m, then output the IV level alarm signal, if it is less than or equal to 15 ·m, then continue to determine whether its minimum value is greater than 10 m, if greater than 10 ·m, then output the level III alarm signal, if it is less than or equal to 10 m, then continue to determine whether its minimum value is greater than 5 m, if greater than 5 ·m, then output the level II alarm signal, if it is less than or equal to 5 ·m, then the level I alarm signal is output.

[0044] Specifically, different preventive measures need to be taken for different levels of alarm signals. Level I alarm signals indicate that the inspection point is seriously harmful and may cause heavy casualties and huge economic losses. The most stringent management measures must be implemented to immediately block the inspection point. Mining operations are prohibited before the risk is eliminated to ensure zero accidents. Level II alarm signals indicate that the inspection point is highly harmful and may cause heavy casualties and major economic losses. Effective management measures must be taken to prohibit workers from going to the inspection point. Level III alarm signals indicate that the inspection point is harmful to a certain extent and may cause general casualties and general economic losses. Supervision and safety precautions need to be strengthened, and workers are prohibited from staying here. Level IV alarm signals are less harmful and may cause minor casualties and minor economic losses. Publicity, education and daily management need to be strengthened to ensure the safety of mining operations.

[0045] In this embodiment, the generated danger level signal is sent to the inspection task analysis platform of the mine dispatching center through the domain controller and displayed on the inspection task analysis platform of the mine dispatching center.

[0046] In one embodiment, the intelligent inspection unmanned vehicle includes a lifting device disposed on the top of the intelligent inspection unmanned vehicle, and also includes a water penetration detection module disposed on the top of the lifting device. The intelligent inspection unmanned vehicle performs water penetration detection on the current target inspection point to obtain a preliminary water penetration detection result. Combining environmental data and the preliminary water penetration detection result to analyze and obtain the water penetration hazard result of the current target inspection point includes the following steps: The lifting device is controlled to lift the water permeability detection module to the top of the mine cave at the current target inspection point for detection, and the transient electromagnetic instrument coil in the water permeability detection module is used to detect the mine cave rock mass at the top of the mine cave to obtain preliminary detection data.

[0047] The preliminary detection data is analyzed to obtain the resistivity value of the mine rock mass at the current target inspection point.

[0048] Analyze the resistivity value to obtain the preliminary water penetration detection results of the current target inspection point.

[0049] The environmental data and preliminary water seepage detection results are analyzed to obtain the water seepage hazard results of the current target inspection point.

[0050] In this embodiment, the water permeability detection module mainly includes a water detection device, and the intelligent patrol unmanned vehicle contains a wire-controlled chassis, a lifting device, a water detection device and other equipment. Among them, the wire-controlled chassis is mainly used to receive signals and control the lifting device. The lifting device is an elevator, which is mainly to facilitate the water detection device to detect the top of the mine cave. The water detection device includes a transient electromagnetic instrument coil, which is used to detect electromagnetic signals inside the rock mass of the mine cave. The equipment connected to the intelligent patrol unmanned vehicle for communication mainly includes a domain controller and an edge computing unit. Among them, the domain controller is mainly used to enhance network security and can save the detection results locally when there is no network. The edge computing unit is mainly used to calculate the internal resistivity of the rock mass.

[0051] Specifically, the intelligent inspection unmanned vehicle used in this embodiment also has a power sensing function, and can send power information to the inspection task analysis platform of the mine dispatching center every 2 hours. The inspection task analysis platform of the mine dispatching center determines whether the battery needs to be replaced. At the same time, according to the power status and task requirements, the inspection task analysis platform of the mine dispatching center sends instructions to the intelligent inspection unmanned vehicle on whether to turn off sensors that do not need to work. When the intelligent inspection unmanned vehicle fails, it can also send fault alarm information to the inspection task analysis platform of the mine dispatching center. After receiving the fault alarm information, the relevant department will dispatch operation and maintenance personnel to perform maintenance work.

[0052] In this embodiment, a start command is sent to the wire-controlled chassis of the intelligent inspection unmanned vehicle through the domain controller, and then the wire-controlled chassis is used to control the lifting device to lift the elevator. After the elevator is raised by the lifting device, the elevator startup result is fed back to the wire-controlled chassis. After receiving the elevator startup result through the wire-controlled chassis, the startup result is sent to the domain controller. After receiving the signal that the lifting device has started, the domain controller sends an instruction to start the water detection device to the edge computing unit. After receiving the instruction to start the water detection device, the edge computing unit controls the intelligent inspection unmanned vehicle to start the water detection device for water detection and obtain preliminary detection data, which is an electromagnetic signal.

[0053] Specifically, the electromagnetic signal is collected mainly by using the transient electromagnetic instrument coil in the water detection device, and a current of a certain waveform is connected to the transient electromagnetic instrument coil, thereby generating an electromagnetic field around the transient electromagnetic instrument coil, and at the same time generating an induced current inside the rock mass of the mine cave, and then the power is cut off. Since the induced current decays over time due to heat loss, the induced electromagnetic field response of the rock mass is measured by the receiving coil to obtain the electromagnetic signal. In addition, the material of the transient electromagnetic instrument coil used in this embodiment is high-density foam, with a size of 90*90cm and a thickness of 3cm.

[0054] In this embodiment, the transient electromagnetic instrument coil in the water detection device is used to collect electromagnetic signals, which are then converted into electromagnetic data through the water detection device and sent to the edge computing unit. The edge computing unit calculates the internal resistivity of the rock mass through the electromagnetic data and the resistivity formula. After the calculation is completed, the environmental data and resistivity data are judged by the water seepage hazard sign identification system inside the edge computing unit to obtain the preliminary water seepage hazard results of the current target inspection point. Then, combined with the environmental data detected by the intelligent unmanned inspection vehicle, the water seepage hazard results of the current target inspection point are analyzed.

[0055] In one embodiment, the environmental data includes environmental humidity data, and combining the environmental data with the preliminary water permeability detection result to obtain the water permeability hazard result of the current target inspection point includes the following steps: A primary mine hidden danger identification model is constructed based on the RBF neural network model.

[0056] Obtain historical environmental humidity data and historical water penetration detection results.

[0057] The historical environmental humidity data and the historical water permeability detection results are input into the primary mine cave hidden danger identification model, the primary mine cave hidden danger identification model is trained, and the mine cave hidden danger identification model is obtained.

[0058] The environmental humidity data and the preliminary water seepage detection results are input into the trained mine hazard identification model, and the environmental humidity data and the preliminary water seepage detection results are mapped to the latent space to obtain the latent space information.

[0059] Based on latent space information, the water seepage hazard results of the target inspection points are output through the mine hazard identification model.

[0060] In this embodiment, the first step in building a primary mine hidden danger identification model based on the RBF neural network model is to obtain historical environmental humidity data and historical water permeability detection results. These historical data may cover different time periods and the amount of data is large, so the data needs to be preprocessed to ensure its quality and consistency. The preprocessing process includes removing noise data, filling missing values, standardization, etc., with the aim of improving the accuracy and reliability of the data. For example, if a sensor fails within a certain period of time, resulting in missing data, the missing values ​​can be filled by interpolation or a prediction model based on historical data.

[0061] Next, these historical data are input into the primary mine hazard identification model and the model is trained. The model used is based on the RBF (Radial Basis Function) neural network. This is because the RBF neural network has good performance in dealing with nonlinear problems, especially for complex scenarios such as mine hazard identification. The training process first needs to design the structure of the RBF neural network, including the input layer, hidden layer, and output layer. The input layer receives historical environmental humidity data and water permeability detection results. The hidden layer maps the input data to a high-dimensional space through the RBF function, and the output layer gives the result of hidden danger identification according to the activation value of the hidden layer. During the training process, the RBF neural network adjusts the center and width of the RBF function in the hidden layer and the weight parameters of the output layer so that the model can accurately capture the complex nonlinear relationship between environmental humidity and water permeability detection results. The goal of the training is to minimize the error between the model prediction value and the actual hidden danger label value. The mean square error (MSE) is usually used as the loss function.

[0062] Through repeated iterative training, the model can gradually learn and identify key features in historical data, including the time series change trend of environmental humidity, the difference between water penetration detection results and preset thresholds, and the correlation between environmental humidity and water penetration detection. Among them, the humidity data is decomposed into three parts: trend, seasonality and residual through the time series decomposition module in the mine hidden danger identification model. The trend part reflects the long-term change trend of humidity data, the seasonal part reflects the periodic fluctuation, and the residual part represents random fluctuation. The mine hidden danger identification model automatically identifies the correlation pattern between environmental humidity and water penetration detection results by learning a large number of joint feature samples. The hidden layer of the RBF neural network maps the input data to a high-dimensional space through the radial basis function, in which the complex relationship between humidity and water penetration is decoupled and expressed in a simpler form. The model adjusts the parameters so that the output results can reflect the correlation in the input features. For example, a rapid increase in humidity may cause abnormal changes in water penetration detection values. Finally, the fully trained primary mine hidden danger identification model can effectively identify potential hidden dangers in mines with high accuracy and reliability.

[0063] After the training is completed, the current environmental humidity data and preliminary water penetration detection results are input into the trained mine cave hidden danger identification model. First, the current environmental humidity data and water penetration detection results are input into the input layer of the model, and after a series of nonlinear transformations, they are mapped to the latent space. The latent space can be understood as a high-dimensional feature space, in which the complex features of the original input data are decoupled and recombined to better reflect the potential relationship between the data. Specifically, the RBF neurons in the hidden layer convert the environmental humidity data and water penetration detection results into a multidimensional vector through the radial basis function. The values ​​on each dimension of the multidimensional vector represent the time series change trend of environmental humidity, the difference between the water penetration detection results and the preset threshold, and the correlation between environmental humidity and water penetration detection. In this way, the latent space can capture the nonlinear characteristics of the input data, so that the model has stronger expressiveness and generalization ability when dealing with complex mine cave environments. In the latent space, the environmental humidity data and water penetration detection results are re-represented as latent space information with stronger recognition ability, providing rich feature information for the next step of hidden danger identification.

[0064] Finally, based on the latent space information obtained, the water seepage hazard results of the target inspection point are output through the mine hazard identification model. The latent space information is passed to the output layer of the model, which converts these latent space information into one or more numerical values ​​through weighted summation and activation function. These values ​​represent the risk level of water seepage hazards in the mine under the current environment. According to the design of the model, the output result can be a binary classification problem (such as the presence or absence of hidden dangers), a multi-classification problem (such as mild hidden dangers, moderate hidden dangers, severe hidden dangers), or a continuous risk score. For example, if the output result is a risk score, then the higher the score, the greater the water seepage hazard at the current inspection point, and vice versa.

[0065] In one embodiment, the environmental data also includes harmful gas concentration data, and judging whether a dangerous sign appears at the current target inspection point according to the water seepage hazard result and the environmental data includes the following steps: If the water seepage hazard result is that there is a water seepage hazard, it is determined that the current target inspection point has dangerous signs.

[0066] If the result of the water seepage hazard is that there is no water seepage hazard, then determine whether there are any dangerous signs at the current target inspection point based on the harmful gas concentration data.

[0067] If the harmful gas concentration data is less than the preset harmful gas concentration threshold, it is determined that there are no dangerous signs at the current target inspection point.

[0068] If the harmful gas concentration data is greater than or equal to the preset harmful gas concentration threshold, it is determined that the current target inspection point has dangerous signs.

[0069] In this embodiment, it is first determined whether there is a water seepage hazard. If the output result of the mine hazard identification model shows that there is a water seepage hazard, it is determined that the current target inspection point has a dangerous sign. If the output result of the mine hazard identification model shows that there is no water seepage hazard, it is determined whether the current target inspection point has a dangerous sign based on the harmful gas concentration data.

[0070] Specifically, in the process of controlling the intelligent inspection unmanned vehicle to drive to the target inspection point, the gas sensor of the intelligent inspection unmanned vehicle remains on to monitor the mine gas in real time, and this process will obtain gas data. Then the intelligent inspection unmanned vehicle is controlled to transmit the gas data to the edge computing unit, and the gas data is analyzed by the edge computing unit to obtain the concentration of harmful gases. Finally, the poisoning and suffocation danger sign identification system installed in the edge computing unit is used to identify the concentration of harmful gases and obtain the inspection result of the current target inspection point, that is, whether there are danger signs at the current target inspection point.

[0071] In one embodiment, controlling the intelligent inspection unmanned vehicle to move toward the target inspection point along the initial inspection route, collecting environmental data and image data in real time through the intelligent inspection unmanned vehicle during the movement toward the target inspection point, and adjusting the initial inspection route to the target inspection route in combination with the environmental data and the image data includes the following steps: Control the intelligent inspection unmanned vehicle to move to the target inspection point along the initial inspection route.

[0072] When the intelligent inspection unmanned vehicle moves toward the target inspection point, the intelligent inspection unmanned vehicle collects environmental data and image data in real time, and the environmental data includes light intensity data.

[0073] If the light intensity data is greater than a preset light intensity threshold, image recognition technology is used to identify the mine terrain features in the image data, and the initial inspection route is adjusted based on the mine terrain features.

[0074] If the light intensity data is less than or equal to the preset light intensity threshold, the initial inspection route is adjusted to the target inspection route.

[0075] In this embodiment, the light intensity sensor is kept in the on state during the process of controlling the intelligent inspection unmanned vehicle to move to the target inspection point. The light intensity sensor is first used to monitor the light intensity of the mine cave, and then the detected light intensity data is transmitted to the edge computing unit. The edge computing unit is used to analyze the light intensity data to determine whether it is greater than the preset light intensity threshold under the light intensity. If the light intensity of the area is less than or equal to the preset light intensity threshold, it means that the light intensity of the area is too small to perform image acquisition, and the inspection route needs to be adjusted; if the light intensity of the area is greater than the preset light intensity threshold, it means that the light intensity of the area is large enough to perform image acquisition. After receiving the information that can be used for image acquisition, the intelligent inspection unmanned vehicle is controlled to turn on the camera to collect images of the current area. After completing the image acquisition work, the image data is obtained, and finally the image recognition technology is used to identify the terrain features of the mine cave in the image data, and the inspection route of the intelligent inspection unmanned vehicle is adjusted according to the terrain features.

[0076] In this embodiment, the light intensity is detected first because image acquisition requires a sufficiently large light intensity to facilitate more accurate analysis and identification of the acquired image data. This can effectively improve the inspection accuracy and work efficiency of the intelligent inspection unmanned vehicle.

[0077] In one embodiment, using image recognition technology to identify mine cave terrain features in image data and adjusting the initial inspection route based on the mine cave terrain features includes the following steps: Preprocess the image data into a grayscale image; The threshold segmentation algorithm is used to separate the target image area containing the complete mine tunnel entrance from the image background in the grayscale image; Locate the target image area through a local binary algorithm and calculate the area size of the target image area; Obtain the camera parameters of the intelligent inspection unmanned vehicle when collecting image data, and use the on-board laser ranging device of the intelligent inspection unmanned vehicle to measure the actual distance between the intelligent inspection unmanned vehicle and the mine entrance; The actual size of the mine entrance is calculated by combining the area size, camera parameters and actual distance; If the actual size of the hole does not match the size of the intelligent inspection unmanned vehicle, the initial inspection route is adjusted to the target inspection route; If the actual size of the hole matches the size of the intelligent inspection unmanned vehicle, image enhancement processing is performed on the target image area; Combine the threshold segmentation algorithm and the local binary algorithm to identify and analyze the target image area after image enhancement processing, and determine whether there are obstacles inside the target mine cave to which the mine cave entrance belongs that hinder the intelligent inspection unmanned vehicle from moving forward; If there are no obstacles in the target mine cave to which the mine cave entrance belongs that hinder the intelligent inspection unmanned vehicle from moving forward, the intelligent inspection unmanned vehicle is controlled to travel along the initial inspection route; If there are obstacles inside the target mine cave to which the mine cave entrance belongs that hinder the advancement of the intelligent inspection unmanned vehicle, the initial inspection route will be adjusted to the target inspection route.

[0078] In this embodiment, the color image contains three channels, RGB, and the value range of each channel is 0 to 255. During preprocessing, the color image is converted into a grayscale image by weighted averaging. Specifically, appropriate weights (usually 0.299, 0.587, 0.114) are selected to weight the sum of the red, green, and blue values ​​of each pixel to obtain a grayscale value. The grayscale image processed in this way retains the brightness information of the original image and removes the color information, making the image processing algorithm more efficient. In this way, each pixel of the entire image is converted into a corresponding grayscale value, thereby obtaining a grayscale image.

[0079] Based on the grayscale image, the threshold segmentation algorithm is used to separate the target image area from the background. The threshold segmentation algorithm is used to divide the image pixels into foreground and background by selecting a threshold. The specific operation is to traverse each pixel of the grayscale image. If the pixel value is greater than the threshold, it is classified as the foreground (for example, set to white); otherwise it is classified as the background (for example, set to black). Selecting an appropriate threshold is the key, which can usually be determined by histogram analysis. Assuming that the grayscale value range of the grayscale image is 0 to 255, draw the grayscale histogram of the image, observe the location of the peak and valley, and select the valley as the threshold. For example, if the histogram of a grayscale image has an obvious valley at the grayscale value of 120, 120 can be selected as the threshold. After this processing, the grayscale image is binarized, and the foreground (mine cave entrance) is clearly separated from the background.

[0080] The local binary algorithm is a binarization method based on local pixel information, which can effectively process images with uneven illumination. The specific method is: define a local window (such as 3×3 or 5×5) around each pixel of the image, calculate the average grayscale value or median value of the pixels in the window, and then use this value as the local threshold of the pixel. If the pixel value is greater than the local threshold, it is classified as foreground; otherwise, it is classified as background. In this way, the target image area can be accurately located. After positioning, the size of the target area (such as area, perimeter, etc.) is calculated.

[0081] The camera parameters of the intelligent inspection unmanned vehicle when collecting image data are obtained. The camera parameters include focal length, sensor size, resolution, etc. These parameters can be obtained through camera calibration. Then the on-board laser ranging device of the intelligent inspection unmanned vehicle is used to measure the actual distance between the intelligent inspection unmanned vehicle and the mine entrance. The on-board laser ranging device can measure the distance between the unmanned vehicle and the mine entrance in real time. The actual size of the mine entrance is calculated by combining the area size, camera parameters and actual distance. According to the camera imaging principle, the actual size can be calculated by camera parameters and image size. The specific formula is: actual size = (image size × actual distance) / (focal length × pixel size). If the actual size of the entrance does not match the vehicle size of the intelligent inspection unmanned vehicle, the initial inspection route is adjusted to the target inspection route. Assuming that the width of the unmanned vehicle is 2 meters and the height is 1.5 meters, and the calculated width of the entrance is 1.5 meters and the height is 1 meter, the size of the entrance is smaller than the size of the unmanned vehicle. At this time, the inspection route needs to be adjusted to prevent the unmanned vehicle from entering the mine that cannot be passed. When adjusting the inspection route, the path planning algorithm can be used to recalculate the optimal route to avoid narrow mine tunnels and ensure the safety of the unmanned vehicle.

[0082] If the actual size of the cave entrance matches the vehicle size of the intelligent inspection unmanned vehicle, the target image area is enhanced. The purpose of image enhancement is to improve the brightness, contrast and clarity of the image, so that subsequent image recognition is more accurate. Common image enhancement methods include histogram equalization, contrast stretching, filtering, etc. Then, the target image area after image enhancement is identified and analyzed by combining the threshold segmentation algorithm and the local binary algorithm, and it is determined whether there are obstacles that hinder the intelligent inspection unmanned vehicle from moving forward inside the target mine cave to which the mine cave entrance belongs. First, the image after image enhancement is binarized by the threshold segmentation algorithm to obtain the foreground and background; then, the foreground is further refined by the local binary algorithm to extract the detailed information inside the mine cave. By analyzing the morphological features of the foreground area (such as area, shape, boundary, etc.), it can be determined whether there are obstacles. For example, the area of ​​a foreground area is 1000 pixels, the shape is an irregular polygon, and the boundary length is 200 pixels. Combined with these features, it can be determined that the area may be an obstacle.

[0083] If there are no obstacles that hinder the intelligent inspection unmanned vehicle from moving forward inside the target mine cave to which the mine cave entrance belongs, the intelligent inspection unmanned vehicle is controlled to drive along the initial inspection route. At this time, the unmanned vehicle can safely pass through the mine cave and continue to perform the inspection task according to the predetermined inspection route. If there are obstacles that hinder the intelligent inspection unmanned vehicle from moving forward inside the target mine cave to which the mine cave entrance belongs, the initial inspection route is adjusted to the target inspection route. At this time, it is necessary to replan the inspection route to avoid obstacles and ensure the driving safety of the unmanned vehicle. The path planning algorithm can use the A* algorithm, the Dijkstra algorithm, etc., to generate a new inspection route by calculating the shortest path or the optimal path. For example, using the A* algorithm, obstacles and drivable areas can be marked on the environmental map, and the optimal path of the unmanned vehicle from the starting point to the end point can be calculated through the heuristic function to generate a new target inspection route. In this way, the unmanned vehicle can avoid obstacles and complete the inspection task safely.

[0084] In another embodiment, using image recognition technology to identify mine cave terrain features in image data, and adjusting the initial inspection route based on the mine cave terrain features includes the following steps: A primary mine geo-identification model is constructed using deep learning models and image recognition technology.

[0085] Input historical image information, train the primary mine cave geographic recognition model, and obtain the mine cave geographic recognition model.

[0086] The mine cave geographic recognition model is used to extract geographic features in the image information, classify and identify the geographic features, and obtain the size characteristics of the mine cave, which include the width and height of the mine cave.

[0087] If the width and height of the mine tunnel are both less than or equal to the height and width of the intelligent inspection unmanned vehicle, the initial inspection route will be adjusted to the target inspection route.

[0088] If the width and height of the mine tunnel are greater than the height and width of the intelligent inspection unmanned vehicle, the mine tunnel geographic recognition model will continue to be used to analyze the geographical features to determine whether there are obstacles in the mine tunnel.

[0089] If there are no obstacles in the mine, the intelligent inspection unmanned vehicle is controlled to travel along the initial inspection route; If there are obstacles in the mine, the initial inspection route will be adjusted to the target inspection route.

[0090] In this embodiment, the deep learning model used is a convolutional neural network model, which is a basic neural network structure mainly used for image recognition and processing, and mainly includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer. Using the convolutional neural network model and image recognition technology to build and train a mine cave geographic recognition model includes the following steps: Step 3.1: Input the historical image information into the convolutional neural network model through the input layer. Since the image information consists of three color channels (red, green, and blue), the image information is converted into a two-dimensional matrix to represent the intensity value of the pixel; Step 3.2: The convolution layer performs a convolution operation on the input historical image information and the convolution kernel, and uses the convolution operation to extract the mine image features in the historical image information. The convolution kernel is equivalent to a small matrix, and this small matrix is ​​convolved with each part of the image. The convolution operation formula is: , in, is the input image, is the convolution kernel, is the position of the convolution kernel on the image, and is the matrix coordinate for convolution operation in the matrix; In this embodiment, the convolution layer includes a first convolution layer and a second convolution layer of two different sizes. The first convolution layer and the second convolution layer respectively contain convolution kernels of different sizes. The first convolution layer is used to extract the mine size features in the historical image information, and the second convolution layer is used to extract the obstacle features in the historical image information.

[0091] Step 3.3: Use maximum pooling or average pooling to reduce the dimension of image information while retaining useful information. Among them, maximum pooling mainly divides the two-dimensional matrix converted by the mine size characteristics and / or obstacle characteristics into multiple 2*2 matrices, selects the maximum value from the values ​​of each 2*2 matrix, and combines them into a new matrix as the output value to retain the most significant mine characteristics. Average pooling is to average the values ​​in the divided 2*2 matrix to obtain the average value, and combine these average values ​​into a new matrix as the mine image pooling feature, thereby obtaining the average representation of the overall feature.

[0092] Step 3.4: Perform full connection processing on the mine cave image pooling features obtained in step 3.3, that is, convert the output value from the two-dimensional mine cave image pooling features into a one-dimensional mine cave image vector feature, which is used to classify the features of the image information; Step 3.5: Convert the one-dimensional mine cave image vector features obtained by full connection processing into specific output results, that is, output the size of the mine cave and whether there are obstacles. Compare the output results with the expected results. If the error is large, perform reverse output and adjust the connection weights between layers until the ideal result is obtained, thereby completing the construction and training of the mine cave geographic recognition model.

[0093] In this embodiment, the collected image information is input into the trained mine cave geographic recognition model to obtain the size characteristics of the mine cave. If the size of the mine cave cannot accommodate the smooth passage of the intelligent inspection unmanned vehicle, the intelligent inspection unmanned vehicle is controlled to re-plan the route; if the size of the mine cave can accommodate the smooth passage of the intelligent inspection unmanned vehicle, the mine cave geographic recognition model is continued to be used to identify whether there are obstacles inside the mine cave. If there are no obstacles, the intelligent inspection unmanned vehicle is controlled to travel according to the initial inspection route; if there are obstacles in the mine cave, the inspection route needs to be re-planned to avoid obstacles and ensure the driving safety of the unmanned vehicle. The path planning algorithm can use the A* algorithm, the Dijkstra algorithm, etc., and generate a new inspection route by calculating the shortest path or the optimal path. For example, using the A* algorithm, obstacles and drivable areas can be marked on the environmental map, and the optimal path of the unmanned vehicle from the starting point to the end point can be calculated by the heuristic function to generate a new target inspection route. In this way, the unmanned vehicle can avoid obstacles and complete the inspection task safely.

[0094] In one embodiment, the environmental data also includes humidity data and corrosive gas concentration data, and combining the environmental data and the image data to determine whether the current target inspection point meets the detection conditions includes the following steps: The dust image is extracted from the image data using image recognition technology.

[0095] The saturation and brightness of the dust image are calculated to obtain the dust image transmittance.

[0096] The dust image transmittance is analyzed and processed to obtain the dust concentration characteristics.

[0097] Combined with dust concentration characteristics, humidity data and corrosive gas concentration data, the kernel principal component analysis method is used to comprehensively analyze whether the current target inspection point meets the detection conditions.

[0098] In this embodiment, after the intelligent inspection unmanned vehicle is controlled to turn on the fill light, the camera is used to collect the dust image of the current target inspection point, and the dust image information is analyzed using image recognition technology. Calculating the saturation and brightness of the dust image to obtain the dust image transmittance includes the following steps: Step 4.1: Describe the mine environment using the atmospheric scattering model: , in, represents a spatial coordinate in the image, represents the dust image collected by the camera, Indicates the light intensity, Indicates light transmittance, is the direct attenuation term.

[0099] Step 4.2: According to the dark channel theory, a color image is divided into three color channels: R, G, and B. When the color of the scene captured by the camera is relatively dim, the image formed has at least one minimum value close to 0 in the three color channels of the color space. The formula for calculating the dark channel value of the dust image is as follows: , in, is the dark channel value at position x in the dust image, Represents the value of color channel c in the local micro-element area y centered at the x pixel position, represents a window centered at pixel x, is the minimum operator.

[0100] Step 4.3: Analyze the formulas in step 4.1 and step 4.2 to obtain the primary formula for transmittance: , in, for The color channels, for The color channel is affected by light intensity, so the transmittance formula is not accurate, so further analysis is required.

[0101] Step 4.4: Definition The dark channel is: , in, for The dark channel of the intermediate formula Indicates that y belongs to Take the minimum value under the condition of Indicates the minimum value among the three color channels, and the same applies to the others.

[0102] Step 4.5: In the HSV color space, hue H represents different colors, saturation S represents the purity of the color, and brightness V represents the brightness of the color. Saturation is related to the purity of a certain hue. Pure spectral colors are fully saturated. With the addition of white light, the saturation gradually decreases. For any three RGB values ​​normalized to the range of [0, 1], the S component and the V component can be calculated by the following formula, that is, the formula for calculating the saturation S and brightness V of the dust image is as follows: , , is the saturation at position x, is the brightness at position x.

[0103] Step 4.6: Substitute the two formulas in step 4.5 into the formula in step 4.4 to obtain: .

[0104] Step 4.7: To prevent the transmittance from being underestimated, It is approximately estimated to be 0. According to the formulas in step 4.6 and step 4.3, the formula for transmittance can be obtained as follows: .

[0105] The dust image is extracted from the image data using image recognition technology, and then the formulas for saturation S and brightness V in step 4.5 are used. Next, the calculated saturation S and brightness V are substituted into the transmittance formula to calculate the transmittance of the dust. Transmittance refers to the change in intensity of light when it passes through a medium, usually expressed as a percentage. The higher the dust concentration, the lower the transmittance, because dust particles block and scatter light. Then, the transmittance is analyzed again using the edge computing unit to obtain the initial dust concentration characteristics. This step is to compare and analyze the transmittance with the known dust concentration model. The dust concentration model is usually a mathematical model based on laboratory experiments and field data, which can describe the change in transmittance at different concentrations. Through this comparative analysis, a preliminary estimate of the dust concentration in the air can be made. In this process, the edge computing unit once again plays its advantage of fast processing, and can complete the comparison and calculation of a large amount of data in a short time. In this way, a preliminary estimate of the dust concentration can be obtained in real time.

[0106] Since the ambient humidity has a significant impact on the dust concentration in the air, the initial dust concentration characteristics need to be corrected. Humidity can affect the hygroscopicity and aggregation of dust particles, thereby changing their distribution and optical properties in the air. Humidity data is usually expressed as a percentage of relative humidity. Next, the humidity data is used to correct the initial dust concentration. The correction process can be achieved by establishing a humidity correction model, in which the effect of humidity on dust concentration can be expressed by a correction coefficient, which can be dynamically adjusted according to the change in humidity. Finally, by combining the initial dust concentration characteristics and the humidity correction coefficient, the final dust concentration characteristics can be calculated. The implementation effect of this process is significant. By considering the influence of humidity, the accuracy and reliability of dust concentration measurement can be greatly improved. After that, the dust concentration characteristics, humidity data and corrosive gas concentration data are combined, and the nuclear principal component analysis method is used to comprehensively analyze whether the current target inspection point meets the detection conditions.

[0107] In one embodiment, combining dust concentration characteristics, humidity data and corrosive gas concentration data, and comprehensively analyzing whether the current target inspection point meets the detection conditions through the kernel principal component analysis method includes the following steps: Map all dust concentration features, humidity data, and corrosive gas concentration data into the same high-dimensional feature space; The principal component analysis method is used to extract fusion data features from the high-dimensional feature space; Construct a mine dangerous environment identification model based on BP neural network model; Input the fused data features into a preset mine dangerous environment identification model, and output the dangerous environment identification results through the mine dangerous environment identification model; According to the dangerous environment identification result, it is judged whether the current environment of the intelligent inspection unmanned vehicle has a corrosive effect on the electronic equipment of the intelligent inspection unmanned vehicle; If the environment in which the intelligent inspection unmanned vehicle is located does not have a corrosive effect on the electronic equipment of the intelligent inspection unmanned vehicle, it is determined that the current target inspection point meets the detection conditions; If the environment in which the intelligent inspection unmanned vehicle is located has a corrosive effect on the electronic equipment of the intelligent inspection unmanned vehicle, it is determined that the current target inspection point does not meet the detection conditions.

[0108] In this embodiment, the above technical solution can be specifically expanded into the following steps: map all dust concentration features, humidity data and corrosive gas concentration data to the same high-dimensional feature space; use principal component analysis in the high-dimensional feature space to extract the first associated data feature between the dust concentration feature and the humidity data, and extract the second associated data feature between the humidity data and the corrosive gas concentration data. Specifically, perform an orthogonal transformation on the dust concentration feature and humidity data in the high-dimensional feature space to extract the first principal component, and filter the first principal component according to a preset condition threshold to obtain the first associated data feature; perform an orthogonal transformation on the humidity data and the corrosive gas concentration data in the high-dimensional feature space to obtain the second principal component, and filter the second principal component according to a preset condition threshold to obtain the second associated data feature; perform feature fusion of the first associated data feature and the second associated data feature with the humidity data as the center to obtain a fused data feature; A mine cave hazardous environment identification model is constructed based on the BP neural network model. The mine cave hazardous environment identification model includes a global feature identification module and a local feature identification module constructed based on the BP neural network. The global feature identification module is used to identify the corrosion effect of the overall data feature of the fused data feature (that is, the common influence of the three data of dust concentration feature, humidity data and corrosive gas concentration data) on the electronic equipment, and the local feature identification module is used to identify the corrosion effect of the local data feature (that is, the first associated data feature and the second associated data feature) in the fused data feature on the electronic equipment. The fused data features are input into the preset mine hazardous environment identification model, and the hazardous environment identification results are output through the mine hazardous environment identification model; based on the hazardous environment identification results, it is determined whether the current environment of the intelligent inspection unmanned vehicle has a corrosive effect on the electronic equipment of the intelligent inspection unmanned vehicle; if the environment of the intelligent inspection unmanned vehicle does not have a corrosive effect on the electronic equipment of the intelligent inspection unmanned vehicle, then it is determined that the current target inspection point meets the detection conditions; if the environment of the intelligent inspection unmanned vehicle has a corrosive effect on the electronic equipment of the intelligent inspection unmanned vehicle, then it is determined that the current target inspection point does not meet the detection conditions.

[0109] In this embodiment, when dust concentration features, humidity data, and corrosive gas concentration data are all mapped to the same high-dimensional feature space, it is necessary to convert these data into a common space for subsequent processing and analysis. First, preprocess each data set to ensure that the dimensions of the data are consistent. This usually includes normalizing or standardizing the data to eliminate the dimensional differences between different features. Normalization can use maximum and minimum value normalization to map the data to the [0,1] interval; standardization is to convert the data into a standard normal distribution with a mean of 0 and a standard deviation of 1. Next, select a suitable kernel function to perform a nonlinear transformation on these data and map them to a high-dimensional feature space. Commonly used kernel functions include Gaussian kernel functions, linear kernel functions, and polynomial kernel functions. The Gaussian kernel function can map data to an infinite-dimensional space, making linearly inseparable data separable in a high-dimensional space. The linear kernel function is suitable for situations where the data is already linearly separable in the original space; the polynomial kernel function can capture high-order nonlinear relationships in the data. By selecting the kernel function and adjusting the parameters, the dust concentration characteristics, humidity data and corrosion gas concentration data are mapped into the same high-dimensional feature space, laying the foundation for subsequent feature extraction and analysis.

[0110] When the principal component analysis method is used in the high-dimensional feature space to extract the first associated data feature between the dust concentration feature and the humidity data, and to extract the second associated data feature between the humidity data and the corrosive gas concentration data, the principal component analysis (PCA) needs to be performed first. The core idea of ​​PCA is to project the original data into a new coordinate system through linear transformation so that the variance of the projected data in this new coordinate system is maximized. In specific implementation, since the ambient humidity will affect both the dust concentration and the gas concentration in the same environment, the principal component analysis is first performed on the dust concentration and humidity data. The covariance matrix of the two variables is calculated, and then the covariance matrix is ​​subjected to eigenvalue decomposition to obtain eigenvalues ​​and eigenvectors. The eigenvector represents the direction of the new coordinate axis, and the eigenvalue represents the variance in these directions. The eigenvector corresponding to the maximum eigenvalue is selected as the first principal component, that is, the first associated data feature between the dust concentration and the humidity data. Similarly, the principal component analysis is performed on the humidity data and the corrosive gas concentration data to obtain the second principal component, that is, the second associated data feature between the humidity data and the corrosive gas concentration data. Through this method, complex high-dimensional data can be reduced to low-dimensional associated features, retaining the main information in the data.

[0111] The first associated data feature and the second associated data feature are fused with humidity data as the center to obtain the fused data feature. The feature fusion method is used to integrate the associated features from different sources. First, the fusion is centered on humidity data because humidity data plays a key role in both associated features. In specific implementation, the first associated data feature and the second associated data feature can be combined by weighted average, series connection or other mathematical operations. For example, the weighted average of the first associated data feature and the second associated data feature can be calculated, where the setting of the weight can be determined based on the importance of each feature or its performance in the training data set. By adjusting the weight parameters, the performance of the fused feature can be optimized to better reflect the comprehensive relationship between dust concentration, humidity and corrosive gas concentration. This feature fusion method can retain the key information of each feature, eliminate the influence of noise data, and improve the robustness and accuracy of the model.

[0112] When building a mine cave hazardous environment identification model based on the BP neural network model, the structure of the neural network needs to be designed first. The BP (Backpropagation) neural network is a feedforward neural network that trains the weights in the network through the backpropagation algorithm. The BP neural network model includes a global feature recognition module and a local feature recognition module. The global feature recognition module is used to process the fused data features, which reflect the overall corrosion effect of dust concentration, humidity and corrosive gas concentration on electronic equipment. In specific implementation, a multi-layer perceptron (MLP) can be designed, each layer containing multiple neurons, and nonlinear transformation is performed through activation functions (such as ReLU, Sigmoid or Tanh). The local feature recognition module is used to process the first associated data features and the second associated data features to capture the detailed relationship between dust concentration and humidity, and humidity and corrosive gas concentration. The local feature recognition module can be implemented using a convolutional neural network (CNN) or a recurrent neural network (RNN) to capture the spatiotemporal features or sequence features in the data. Through the combination of global and local feature recognition modules, the BP neural network model can more accurately predict the corrosion risk of the mine cave environment to the electronic equipment of the intelligent inspection unmanned vehicle, and improve the prediction ability and robustness of the model. The fused data features are input into the preset mine dangerous environment identification model. When the mine dangerous environment identification model outputs the dangerous environment identification result, the fused data features obtained in the previous step need to be used as the input of the model. The model outputs a dangerous environment identification result through the processing of the global feature recognition module and the local feature recognition module. This result can be a binary classification (such as dangerous and not dangerous), a multivariate classification (such as slight corrosion, moderate corrosion and severe corrosion), or a continuous value (such as corrosion index).

[0113] When judging whether the current environment of the intelligent inspection unmanned vehicle has a corrosive effect on the electronic equipment of the intelligent inspection unmanned vehicle based on the results of dangerous environment identification, it is necessary to make corresponding judgments based on the prediction results of the model. If the identification results show that the current environment does not have a corrosive effect, it means that the environmental conditions of the inspection point are good and meet the detection conditions, and no additional measures are required. If the results show that there is a risk of erosion in the environment, it means that the conditions of the inspection point do not meet the requirements and corresponding measures need to be taken. For example, the inspection route can be changed to avoid high-risk areas; the inspection plan can be adjusted to increase the monitoring frequency of high-risk areas; or additional environmental protection operations can be performed, such as adding protective measures and improving ventilation conditions. In this way, the corrosion risk of electronic equipment can be effectively reduced, the service life of the equipment can be extended, and the inspection efficiency can be improved. The overall solution reduces the error of human judgment through the intelligent judgment of the model, ensures the scientificity and accuracy of the inspection process, and improves the application effect of the intelligent inspection unmanned vehicle in the mine environment.

[0114] The present invention also discloses an intelligent inspection unmanned vehicle, comprising: A base, with a moving device for movement arranged at the bottom; A lifting device is arranged on the top of the base; The water permeability detection module is installed on the top of the lifting device and is used to measure the resistivity value of the rock mass in the mine; Gas sensors are installed on both sides of the base to detect the concentration data of harmful gases in the surrounding environment; Humidity sensors are arranged on both sides of the base to detect ambient humidity data of the surrounding environment; Light sensors are arranged on both sides of the base to detect the light intensity data of the surrounding environment; A high-definition camera is installed in front of the intelligent inspection unmanned vehicle to collect image data of the surrounding environment.

[0115] An embodiment of the present application also provides a machine-readable storage medium, on which instructions are stored, and the instructions are used to enable a machine to execute the above-mentioned non-coal mine intelligent inspection method.

[0116] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0117] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0118] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1A function specified in one or more boxes.

[0119] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0120] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0121] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0122] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0123] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0124] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A non-coal mine intelligent inspection method, characterized in that: Applied to an intelligent inspection unmanned vehicle, the method comprises the following steps: Obtaining a mine area map of a target mine area, wherein the mine area map is pre-marked with a plurality of dangerous areas; Determine a plurality of target inspection points according to the plurality of dangerous areas, and generate an initial inspection route based on all the target inspection points; Controlling the intelligent inspection unmanned vehicle to move toward the target inspection point along the initial inspection route; during the movement of the intelligent inspection unmanned vehicle toward the target inspection point, collecting environmental data and image data of the surrounding environment in real time through the intelligent inspection unmanned vehicle, and adjusting the initial inspection route to the target inspection route in combination with the environmental data and the image data; Controlling the intelligent inspection unmanned vehicle to move toward the target inspection point according to the target inspection route; After arriving at any of the target inspection points, judging whether the current target inspection point meets the detection conditions by combining the environmental data and the image data; If the current target inspection point meets the detection conditions, the intelligent inspection unmanned vehicle performs water seepage detection on the current target inspection point to obtain a preliminary water seepage detection result, and combines the environmental data and the preliminary water seepage detection result to obtain a water seepage hazard result of the current target inspection point; Determine whether the target inspection point currently has any danger signs according to the water seepage hazard result and the environmental data; The environmental data and the water penetration detection result are combined to analyze and determine the current danger level of the target inspection point.

2. The method according to claim 1, characterized in that: The intelligent inspection unmanned vehicle includes a lifting device disposed on the top of the intelligent inspection unmanned vehicle, and also includes a water penetration detection module disposed on the top of the lifting device. The intelligent inspection unmanned vehicle performs water penetration detection on the current target inspection point to obtain a preliminary water penetration detection result, and combines the environmental data and the preliminary water penetration detection result to analyze and obtain the water penetration hazard result of the current target inspection point, including the following steps: Control the lifting device to lift the water permeability detection module to the top of the mine cave at the current target inspection point, and use the transient electromagnetic instrument coil in the water permeability detection module to detect the rock mass at the top of the mine cave to obtain preliminary detection data; Analyze the preliminary detection data to obtain the resistivity value of the rock mass in the mine at the current target inspection point; Analyze the resistivity value to obtain the preliminary water penetration detection result of the current target inspection point; The environmental data and the preliminary water seepage detection result are analyzed in combination to obtain the water seepage hazard result of the current target inspection point.

3. The method according to claim 2, characterized in that The environmental data includes environmental humidity data, and the step of combining the environmental data with the preliminary water permeability detection result to obtain the water permeability hazard result of the current target inspection point includes the following steps: Construct a primary mine hidden danger identification model based on RBF neural network model; Obtain historical environmental humidity data and historical water penetration detection results; Inputting the historical environmental humidity data and the historical water permeability detection results into the primary mine cave hidden danger identification model, training the primary mine cave hidden danger identification model, and obtaining a mine cave hidden danger identification model; Inputting the environmental humidity data and the preliminary water permeability detection result into the trained mine hazard identification model, mapping the environmental humidity data and the preliminary water permeability detection result into a latent space to obtain latent space information; Based on the latent space information and through the mine hazard identification model, the water seepage hazard result of the target inspection point is output.

4. The method according to claim 1, characterized in that: The environmental data also includes harmful gas concentration data. The step of judging whether the target inspection point currently has dangerous signs based on the water seepage hazard result and the environmental data includes the following steps: If the water seepage hidden danger result is that there is a water seepage hidden danger, it is determined that the current target inspection point has a dangerous sign; If the water seepage hidden danger result is that there is no water seepage hidden danger, judging whether there is any danger sign at the current target inspection point according to the harmful gas concentration data; If the harmful gas concentration data is less than the preset harmful gas concentration threshold, it is determined that there is no danger sign at the current target inspection point; If the harmful gas concentration data is greater than or equal to a preset harmful gas concentration threshold, it is determined that a dangerous sign appears at the current target inspection point.

5. The method according to claim 1, characterized in that The controlling the intelligent inspection unmanned vehicle to move toward the target inspection point according to the initial inspection route, collecting environmental data and image data in real time by the intelligent inspection unmanned vehicle during the movement of the intelligent inspection unmanned vehicle toward the target inspection point, and adjusting the initial inspection route to the target inspection route in combination with the environmental data and the image data comprises the following steps: Controlling the intelligent inspection unmanned vehicle to move toward the target inspection point according to the initial inspection route; In the process of the intelligent inspection unmanned vehicle moving toward the target inspection point, the intelligent inspection unmanned vehicle collects environmental data and image data in real time, wherein the environmental data includes light intensity data; If the light intensity data is greater than a preset light intensity threshold, the image recognition technology is used to identify the mine cave terrain features in the image data, and the initial inspection route is adjusted based on the mine cave terrain features; If the light intensity data is less than or equal to a preset light intensity threshold, the initial inspection route is adjusted to a target inspection route.

6. The method according to claim 5, characterized in that The method of using image recognition technology to identify the mine cave terrain features in the image data and adjusting the initial inspection route based on the mine cave terrain features includes the following steps: Preprocessing the image data into a grayscale image; A threshold segmentation algorithm is used to separate the target image region including the complete mine tunnel entrance from the image background in the grayscale image; Locating the target image area by using a local binary algorithm, and calculating the area size of the target image area; Obtaining the camera parameters of the intelligent inspection unmanned vehicle when collecting image data, and using the on-board laser ranging device of the intelligent inspection unmanned vehicle to measure the actual distance between the intelligent inspection unmanned vehicle and the mine entrance; The actual size of the mine entrance is calculated by combining the area size, the camera parameters and the actual distance; If the actual size of the hole does not match the vehicle size of the intelligent inspection unmanned vehicle, the initial inspection route is adjusted to a target inspection route; If the actual size of the hole matches the vehicle size of the intelligent inspection unmanned vehicle, image enhancement processing is performed on the target image area; Combining the threshold segmentation algorithm and the local binary algorithm to identify and analyze the target image area after image enhancement processing, and judging whether there are obstacles inside the target mine cave to which the mine cave entrance belongs that hinder the advancement of the intelligent inspection unmanned vehicle; If there is no obstacle inside the target mine cave to which the mine cave entrance belongs that hinders the advance of the intelligent inspection unmanned vehicle, the intelligent inspection unmanned vehicle is controlled to travel along the initial inspection route; If there are obstacles inside the target mine cave to which the mine cave entrance belongs that hinder the advancement of the intelligent inspection unmanned vehicle, the initial inspection route is adjusted to the target inspection route.

7. The method according to claim 3, characterized in that The environmental data also includes corrosive gas concentration data, and the combination of the environmental data and the image data to determine whether the current target inspection point meets the detection conditions includes the following steps: extracting a dust image from the image data using image recognition technology; Calculating the saturation and brightness of the dust image to obtain the light transmittance of the dust image; Analyzing and processing the light transmittance of the dust image to obtain the dust concentration characteristics; The dust concentration characteristics, the environmental humidity data and the corrosive gas concentration data are combined, and a comprehensive analysis is performed through a kernel principal component analysis method to determine whether the current target inspection point meets the detection conditions.

8. The method according to claim 7, characterized in that The step of combining the dust concentration characteristics, the humidity data and the corrosive gas concentration data and comprehensively analyzing whether the current target inspection point meets the detection conditions by using the kernel principal component analysis method comprises the following steps: Mapping the dust concentration characteristics, the humidity data and the corrosive gas concentration data to the same high-dimensional feature space; Extracting fusion data features from the high-dimensional feature space using principal component analysis; Construct a mine dangerous environment identification model based on BP neural network model; Inputting the fused data features into the mine dangerous environment identification model, and outputting the dangerous environment identification result through the mine dangerous environment identification model; Determining whether the current environment of the intelligent inspection unmanned vehicle has a corrosive effect on the electronic equipment of the intelligent inspection unmanned vehicle according to the dangerous environment identification result; If the environment in which the intelligent inspection unmanned vehicle is located does not have a corrosive effect on the electronic equipment of the intelligent inspection unmanned vehicle, it is determined that the current target inspection point meets the detection conditions; If the environment in which the intelligent inspection unmanned vehicle is located has a corrosive effect on the electronic equipment of the intelligent inspection unmanned vehicle, it is determined that the current target inspection point does not meet the detection conditions.

9. The method according to claim 8, characterized in that The extracting of fusion data features from the high-dimensional feature space by using principal component analysis method comprises the following steps: Performing an orthogonal transformation on the data in the high-dimensional feature space to obtain principal components; The principal components are screened according to a preset condition threshold to obtain the fused data features.

10. An intelligent inspection unmanned vehicle, characterized in that: Applicable to the non-coal mine intelligent inspection method according to any one of claims 1 to 9, the intelligent inspection unmanned vehicle comprises: A base, with a moving device for movement arranged at the bottom; A lifting device, arranged on the top of the base; A water permeability detection module is arranged on the top of the lifting device and is used to measure the resistivity value of the rock mass in the mine; Gas sensors are arranged on both sides of the base to detect the concentration data of harmful gases in the surrounding environment; Humidity sensors are arranged on both sides of the base and are used to detect ambient humidity data of the surrounding environment; Light-sensitive sensors are arranged on both sides of the base and are used to detect light intensity data of the surrounding environment; A high-definition camera is arranged in front of the intelligent inspection unmanned vehicle and is used to collect image data of the surrounding environment.

Citation Information

Patent Citations

  • Autonomous navigation method and system of high-voltage transmission line inspection robot

    CN108007464A

  • Building health monitoring system based on field monitoring as well as monitoring method

    CN109883474A

  • Transient electromagnetic unmanned inspection equipment for underground water permeation monitoring and monitoring method

    CN117192620A

  • Task scheduling and analyzing method for unmanned inspection vehicle in complex mine environment

    CN117634801A

  • Control method of inspection robot, computer program product and inspection system

    CN118244767A