A road tunnel safety hazard detection method, device and medium

Abstract

The embodiment of the present application discloses a method, device and medium for detecting safety hazards in highway tunnels. It belongs to the field of highway tunnel inspection technology and solves the problem that it is difficult to judge the safety hazards of tunnels in a short time. Obtain environmental information and tunnel images in the highway tunnel; determine the correlation between different detection points based on the environmental information, and detect the fire incidence in the highway tunnel based on the correlation, the temperature information corresponding to the different detection points and the preset temperature prediction model; based on the opening and closing operations, amplify and extract the crack features of the tunnel image; based on the pixel grayscale values ​​corresponding to the tunnel image, divide the tunnel image after the crack features are amplified into multiple different connected domains to filter the noise of the tunnel image, and detect the crack conditions of the highway tunnel based on the tunnel image after the noise is filtered; determine the safety hazard situation of the highway tunnel based on the fire incidence and the crack conditions.

Description

Technical Field

[0001] The present application relates to the technical field of highway tunnel inspection, and in particular to a method, equipment and medium for detecting safety hazards in highway tunnels. Background Art

[0002] Tunnel construction is gradually moving towards intelligence and electrification, and at the same time, the tasks of later operation and maintenance are also becoming heavier. Due to the influence of various factors such as construction, temperature, and load, cracks on the surface of tunnels are common, posing a threat to the safe operation of tunnels. Therefore, crack detection is an important task of periodic tunnel inspections. At the same time, tunnel construction has many electrical equipment inside, the power grid is complex, and high-power current transmission also creates a high fire hazard. The later operation and maintenance work of tunnels involves not only engineering aspects such as lining cracks and leakage, but also aspects such as tunnel fire prevention.

[0003] With the increase in tunnel mileage and traffic volume, the types of tunnel equipment and facilities are also gradually increasing, which brings great pressure to the daily maintenance and management of tunnels. At present, tunnel operation inspection and maintenance are mainly manual, which has problems such as long time consumption, low efficiency, easy to miss inspections, and the accuracy of inspection results is limited by the technical level of inspection personnel, making it difficult for tunnel managers to judge the safety hazards of tunnels in a short time. Summary of the invention

[0004] The embodiments of the present application provide a method, device and medium for detecting safety hazards in highway tunnels, which are used to solve the following technical problems: Currently, the operation inspection and maintenance of tunnels are mainly manual, which has the problems of being time-consuming, inefficient, easy to miss detections, and the accuracy of detection results is limited by the technical level of the inspection personnel, making it difficult for tunnel managers to judge the safety hazards of the tunnel in a short period of time.

[0005] The present application embodiment adopts the following technical solutions:

[0006] The embodiment of the present application provides a method for detecting safety hazards in a highway tunnel. The method comprises: obtaining environmental information and tunnel images in the highway tunnel through multiple sensor components arranged at different detection points in the highway tunnel; determining the correlation between different detection points based on the environmental information, and detecting the fire incidence rate in the highway tunnel based on the correlation, the temperature information corresponding to the different detection points, and a preset temperature prediction model; predicting the severity level of the fire based on the fire information obtained by the sensor when the detection result is a fire, and processing the fire based on the prediction result; amplifying and extracting crack features from the tunnel image based on an open operation and a closed operation; dividing the tunnel image after crack feature amplification into multiple different connected domains based on the pixel grayscale value corresponding to the tunnel image, so as to filter the noise of the tunnel image, and detecting crack conditions in the highway tunnel based on the tunnel image after noise filtering; determining the safety hazard conditions of the highway tunnel based on the fire incidence rate and the crack conditions.

[0007] The embodiment of the present application obtains environmental information in the highway tunnel through the sensor assembly set in the highway tunnel, and can analyze the environmental information to determine the safety hazards such as fire in the highway tunnel. Secondly, the embodiment of the present application obtains the tunnel image in the highway tunnel, extracts the crack features of the tunnel image, and performs noise filtering through the gray value of the image pixels, thereby eliminating the protruding burrs on the cracks and making the overall contour smooth, while minimizing the change in the area of ​​the crack connection domain, effectively eliminating the noise in the image. This solves the problems of low efficiency, low accuracy, and low degree of automation in the detection of tunnel safety hazards, and also improves the comprehensiveness and timeliness of the detection.

[0008] In one implementation of the present application, the correlation between different detection points is determined based on environmental information, specifically including: determining the Euclidean distance between different detection points under different environmental parameters; wherein the environmental parameters include at least one of temperature information, flame information and smoke information; establishing a correlation matrix based on the Euclidean distance, and determining a reference correlation between different detection points based on the correlation matrix; obtaining wind direction information and wind speed information in the tunnel in real time, and determining a first reference weight value in a first preset weight database based on the wind direction information and wind speed information; obtaining tunnel layout information between different detection points, and determining a second reference weight value in a second preset weight database based on the tunnel layout information; determining the correlation between different detection points based on the reference correlation, the first reference weight value and the second reference weight value.

[0009] In one implementation of the present application, a fire incidence rate in a highway tunnel is predicted based on a correlation, temperature information corresponding to different detection points, and a preset temperature prediction model, specifically comprising: inputting temperature information acquired by multiple sensors into a preset temperature prediction model to determine predicted temperature information corresponding to multiple detection points in the tunnel at multiple future time points; comparing the predicted temperature information corresponding to multiple future time points with preset reference temperature information to determine a set of temperature difference values ​​corresponding to multiple detection points; wherein the preset reference temperature information includes critical temperatures for fire occurrence corresponding to different times in the tunnel; based on the correlation, determining temperature influence values ​​of multiple detection points on a current detection point; adjusting the predicted temperature information corresponding to the current detection point according to the temperature influence value and the set of temperature difference values ​​corresponding to the multiple detection points, so as to determine the fire incidence rate of the current detection point based on the adjusted predicted temperature information.

[0010] In one implementation of the present application, the severity level of a fire is predicted based on the fire information acquired by the sensor, and the fire is processed based on the prediction result, specifically including: acquiring flame information respectively sent by multiple flame sensors within a preset time period, and determining the direction of the fire in the highway tunnel based on the flame information; acquiring temperature information respectively sent by multiple temperature sensors within a preset time period, and determining the distance between the fire location and the tunnel inspection robot based on the temperature information; acquiring multiple smoke information sent by the smoke sensor within a preset time period, so as to detect the current fire severity level based on the smoke information, flame information and temperature information; based on the detection results, issuing different processing instructions to the tunnel inspection robot, and controlling the tunnel inspection robot to handle the fire based on the direction and distance.

[0011] In one implementation of the present application, based on the pixel grayscale values ​​corresponding to the tunnel image, the tunnel image after the crack feature is amplified is divided into multiple different connected domains to perform noise filtering on the tunnel image, specifically including: obtaining a crack feature amplified image based on the tunnel image after opening and closing operations; dividing the crack feature amplified image into multiple equal-sized regions; dividing the crack feature amplified image into multiple connected regions based on the pixel grayscale values ​​corresponding to the multiple equal-sized regions; determining the areas corresponding to the multiple connected regions, so as to denoise the crack feature image based on the areas corresponding to the multiple connected regions.

[0012] In one implementation of the present application, the crack feature enlarged image is divided into multiple connected areas based on the pixel grayscale values ​​corresponding to multiple equal-sized areas, specifically including: when the pixel grayscale values ​​of the four points on the left, upper left, top and upper right in any equal-sized area are all 0, it is marked as the starting point of a new connected domain; or, when the pixel grayscale values ​​of the two points on the left and upper left in any equal-sized area are all 1, the smaller point of the two pixel marking values ​​on the left and upper left is used as the marking point; or, when the pixel grayscale values ​​of the two points on the upper left and upper right in any equal-sized area are all 1, the smaller point of the two pixels on the upper left and upper right is used as the marking point; or, the marking points are marked in the order of left, upper left, top and upper right; traverse multiple equal-sized areas, and record the areas with the same mark as a connected domain.

[0013] In one implementation of the present application, the areas corresponding to the multiple connected regions are determined, and the crack feature image is denoised based on the areas corresponding to the multiple connected regions, specifically including: determining the areas corresponding to the multiple connected regions, and determining the areas of the minimum circumscribed rectangles corresponding to the multiple connected regions; when the ratio of the area corresponding to the current connected region to the area of ​​the minimum circumscribed rectangle corresponding to the current connected region is greater than the preset connected domain pixel ratio, the current connected region is retained, otherwise the current connected region is used as the background to denoise the crack feature image.

[0014] In one implementation of the present application, before crack detection is performed on a highway tunnel based on a noise-filtered tunnel image, the method further includes: determining a prediction frame accuracy ratio corresponding to the tunnel sample image category based on the number of frames with correct target detection results corresponding to the tunnel sample image category and the total number of all frames detected for the tunnel sample image category; determining an average accuracy rate corresponding to the tunnel sample image category based on the prediction frame accuracy ratio and the number of images corresponding to the tunnel sample image category; and determining an average accuracy rate corresponding to a preset crack detection model based on the number of image categories and the average accuracy rate corresponding to images of each category.

[0015] The embodiment of the present application provides a road tunnel safety hazard detection device, comprising: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can: obtain environmental information and tunnel images in the road tunnel through multiple sensor components arranged at different detection points in the road tunnel; determine the correlation between different detection points based on the environmental information, and detect the fire incidence rate in the road tunnel based on the correlation, temperature information corresponding to the different detection points, and a preset temperature prediction model; when the detection result is that a fire has occurred, predict the severity level of the fire based on the fire information obtained by the sensor, and process the fire based on the prediction result; based on opening and closing operations, extract crack features from the tunnel image; based on the pixel grayscale values ​​corresponding to the tunnel image, divide the tunnel image after the crack features are amplified into multiple different connected domains to filter the tunnel image, and detect crack conditions in the road tunnel based on the tunnel image after the noise is filtered; based on the fire incidence rate and the crack conditions, determine the safety hazard conditions of the road tunnel.

[0016] A non-volatile computer storage medium provided by an embodiment of the present application stores computer executable instructions, wherein the computer executable instructions are configured to: obtain environmental information and a tunnel image in a highway tunnel by means of a plurality of sensor components disposed at different detection points in the highway tunnel; determine the correlation between different detection points based on the environmental information, and detect the fire incidence in the highway tunnel based on the correlation, the temperature information corresponding to the different detection points, and a preset temperature prediction model; when the detection result is that a fire has occurred, predict the severity level of the fire based on the fire information obtained by the sensor, and process the fire based on the prediction result; amplify and extract crack features from the tunnel image based on an opening operation and a closing operation; divide the tunnel image after the crack features are amplified into a plurality of different connected domains based on the pixel grayscale values ​​corresponding to the tunnel image, so as to perform noise filtering on the tunnel image, and detect crack conditions in the highway tunnel based on the tunnel image after the noise filtering; determine the safety hazard of the highway tunnel based on the fire incidence and the crack conditions.

[0017] At least one of the above technical solutions adopted in the embodiments of the present application can achieve the following beneficial effects: the embodiments of the present application obtain environmental information in the highway tunnel through the sensor components arranged in the highway tunnel, and can analyze the environmental information to determine the safety hazards such as fire in the highway tunnel. Secondly, the embodiments of the present application obtain the tunnel image of the highway tunnel, extract the crack features of the tunnel image, and filter the noise through the gray value of the image pixels, so as to eliminate the protruding burrs on the cracks and make the overall contour smooth, while minimizing the change in the area of ​​the crack connection domain, and effectively eliminating the noise in the image. This solves the problem of backward tunnel safety detection, low efficiency and low accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. In the drawings:

[0019] Figure 1 A flow chart of a method for detecting safety hazards in highway tunnels provided in an embodiment of the present application;

[0020] Figure 2 A schematic structural diagram of a highway tunnel safety hazard detection device provided in an embodiment of the present application.

[0021] Reference numerals:

[0022] 200: Highway tunnel safety hazard detection equipment, 201: processor, 202: memory. DETAILED DESCRIPTION

[0023] The embodiments of the present application provide a method, device and medium for detecting safety hazards in a highway tunnel.

[0024] In order to enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of them. Based on the embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this application.

[0025] The technical solution proposed in the embodiments of the present application is described in detail below with reference to the accompanying drawings.

[0026] Figure 1A flow chart of a method for detecting safety hazards in highway tunnels provided in an embodiment of the present application. Figure 1 As shown, the road tunnel safety hazard detection method includes the following steps:

[0027] Step 101: Acquire environmental information and tunnel images in a highway tunnel by using multiple sensor components arranged at different detection points in the highway tunnel.

[0028] In one embodiment of the present application, for obtaining environmental information, a series of environmental monitoring sensors are installed in the tunnel, such as temperature and humidity sensors, smoke sensors, and harmful gas sensors. The installed sensors can monitor the temperature and humidity, smoke concentration, and harmful gas content in the tunnel in real time, providing important data support for the safe operation of the tunnel. Secondly, a camera device is installed in the tunnel to shoot the inner surface of the tunnel, and the crack information in the tunnel is analyzed through the tunnel images obtained by shooting.

[0029] Step 102: Determine the correlation between different detection points based on the environmental information, and detect the fire incidence in the highway tunnel based on the correlation, the temperature information corresponding to the different detection points, and a preset temperature prediction model.

[0030] In one embodiment of the present application, the Euclidean distance between different detection points under different environmental parameters is determined; wherein the environmental parameters include at least one of temperature information, flame information, and smoke information. A correlation matrix is ​​established based on the Euclidean distance, and a reference correlation between different detection points is determined based on the correlation matrix. The wind direction information and wind speed information in the tunnel are acquired in real time, and a first reference weight value is determined in a first preset weight database based on the wind direction information and wind speed information. The tunnel layout information between different detection points is acquired, and a second reference weight value is determined in a second preset weight database based on the tunnel layout information. Based on the reference correlation, the first reference weight value, and the second reference weight value, the correlation between different detection points is determined.

[0031] Specifically, environmental parameter data of each detection point is collected, and the environmental parameter data includes at least one of temperature information, flame information and smoke information. Then, the Euclidean distance between different detection points is calculated. For each environmental parameter, the difference between two detection points can be calculated, and the square root of the sum of the squares of these differences can be taken to obtain the Euclidean distance between them.

[0032] Furthermore, a correlation matrix is ​​established based on the calculated Euclidean distance. Each row and column of the correlation matrix represents a detection point, and the values ​​in the matrix represent the correlation between the corresponding detection points. The correlation can be the reciprocal of the Euclidean distance, or a value obtained by performing other transformations based on the Euclidean distance. Furthermore, from the correlation matrix, by selecting the values ​​of the corresponding positions in the matrix, the reference correlation between different detection points can be directly read or calculated.

[0033] Furthermore, the wind direction and wind speed information in the tunnel is obtained in real time. This information is crucial for determining the direction of fire or smoke diffusion. According to the wind direction and wind speed, the corresponding first reference weight value is searched in the first preset weight database, wherein the first preset weight database contains the mapping relationship between different wind direction and wind speed combinations and weight values. The tunnel layout information includes the length, width, curve position and relative position between detection points of the tunnel. Based on these tunnel layout information, the corresponding second reference weight value is searched in the second preset weight database, wherein the second preset weight database contains the mapping relationship between different tunnel layout features and weight values.

[0034] Furthermore, the final correlation between different detection points is determined by combining the reference correlation, the first reference weight value, and the second reference weight value. Specifically, the reference correlation, the first reference weight value, and the second reference weight value are multiplied, and then the results are added to obtain a comprehensive correlation value, thereby determining the correlation between different detection points.

[0035] In one embodiment of the present application, the temperature information obtained by multiple sensors is respectively input into a preset temperature prediction model to determine the predicted temperature information corresponding to multiple detection points in the tunnel at multiple time points in the future. The predicted temperature information corresponding to multiple time points in the future is compared with the preset reference temperature information to determine the temperature difference sets corresponding to the multiple detection points; wherein the preset reference temperature information includes the critical temperatures for fire occurrence corresponding to different times in the tunnel. Based on the correlation, the temperature influence values ​​of the multiple detection points on the current detection point are determined. According to the temperature influence value and the temperature difference sets corresponding to the multiple detection points, the predicted temperature information corresponding to the current detection point is adjusted to determine the fire incidence rate of the current detection point based on the adjusted predicted temperature information.

[0036] Specifically, multiple detection points are set up in the tunnel, and each detection point is equipped with a sensor for obtaining the temperature information of the point in real time. These sensors will continuously collect data and transmit the data to the central processing system for analysis. In the central processing system, a temperature prediction model is pre-built, which can predict the future temperature based on past temperature data and other possible factors. The training process of the model is to use the historical temperature data of n time periods in the tunnel as input samples, and the historical temperature data corresponding to the n+1th time period as output samples, and train the neural network model to obtain the temperature prediction model. The real-time temperature data obtained by each sensor is input into this model, and the temperature prediction model will output the predicted temperature information of the corresponding detection point at multiple time points in the future (for example, in the next few hours or days).

[0037] Furthermore, the embodiment of the present application is provided with a preset reference temperature information, and the preset reference temperature information includes the critical temperature of fire occurrence corresponding to different times in the tunnel. This critical temperature is determined based on factors such as tunnel materials, ventilation conditions, historical fire data, etc. When the temperature exceeds this critical value, the risk of fire will increase significantly. The predicted temperature of each detection point at multiple time points in the future is compared with the corresponding critical temperature to determine a temperature difference set. The temperature difference set in the embodiment of the present application reflects the gap between each detection point and the critical temperature of fire within the predicted time period.

[0038] Furthermore, since the temperature distribution in the tunnel may be affected by adjacent detection points (e.g., hot air flow, heat transfer, etc.), it is necessary to determine the correlation between the detection points. Based on these correlations, the temperature impact value of each detection point on other detection points can be calculated, that is, the temperature impact value indicates the degree of influence of the temperature change of one detection point on the temperature of another detection point.

[0039] Furthermore, the predicted temperature information of the current detection point is adjusted using the calculated temperature influence value in combination with the temperature difference set of each detection point, so as to more accurately reflect the possible temperature changes of the current detection point in the future. For example, if the weighted temperature influence value is positive, and the temperature difference set shows that the temperature prediction value of the adjacent detection point is higher than that of the current detection point, then it can be considered that the temperature of the current detection point may be affected by the heating effect of the adjacent point, and therefore the predicted temperature of the current detection point should be appropriately increased. Conversely, if the weighted temperature influence value is negative, and the temperature difference set shows that the temperature prediction value of the adjacent detection point is lower than that of the current detection point, then the predicted temperature of the current detection point can be appropriately reduced.

[0040] Step 103: When the detection result indicates that a fire has occurred, the severity of the fire is predicted based on the fire information acquired by the sensor, and the fire is processed based on the prediction result.

[0041] In one embodiment of the present application, flame information sent by multiple flame sensors within a preset time period is obtained, and the position of the fire in the highway tunnel is determined based on the flame information. Temperature information sent by multiple temperature sensors within a preset time period is obtained, and the distance between the fire position and the tunnel inspection robot is determined based on the temperature information. Multiple smoke information sent by the smoke sensor within a preset time period is obtained to detect the current fire severity level based on the smoke information, flame information and temperature information. Based on the detection results, different processing instructions are issued to the tunnel inspection robot, and the tunnel inspection robot is controlled to handle the fire based on the position and distance.

[0042] Specifically, the flame sensor can promptly determine the occurrence of a fire. If multiple flame sensors are used, the direction of the fire can be determined. For example, the direction corresponding to the flame sensor with the most obvious flame data is the direction of the fire, and then the tunnel inspection robot is controlled to move in the direction of the fire. The flame sensor is very sensitive to light signals of a specific wavelength. The sensitivity of the flame sensor can be adjusted by adjusting the adjustable resistor, so that the flame sensor can detect the flame more accurately and is not interfered by other light sources. If a wavelength similar to the flame light is detected, a message is immediately sent to the host computer to issue an alarm message.

[0043] Through the temperature sensor, the temperature conditions in a specific direction and distance can be monitored in real time. By setting a threshold, when the temperature exceeds this value, it means that the temperature is abnormal and it is very likely that a fire has occurred. Then an alarm signal will be issued immediately to trigger the alarm device.

[0044] Specifically, the embodiment of the present application is provided with a first fire detection model, and the training process of the model is to take the sample flame data as input, take the fire level corresponding to the sample flame data as output, and train the preset first neural network model to obtain the first fire detection model. Through the first fire detection model, the flame information corresponding to the current multiple time points can be analyzed and processed to output the fire level corresponding to the flame information. Secondly, the embodiment of the present application is provided with a second fire detection model, and the training process of the model is to take the sample temperature data as input, take the fire level corresponding to the sample temperature data as output, and train the preset second neural network model to obtain the second fire detection model. Through the second fire detection model, the temperature information corresponding to the current multiple time points can be analyzed and processed to output the fire level corresponding to the temperature information. In addition, the embodiment of the present application is provided with a third fire detection model, and the training process of the model is to take the sample smoke information as input, take the fire level corresponding to the sample smoke information as output, and train the preset third neural network model to obtain the third fire detection model. The third fire detection model can be used to analyze and process the smoke information corresponding to the current multiple time points to output the fire level corresponding to the smoke information. The fire level with the largest number of output fire levels is used as the fire level corresponding to the current highway tunnel.

[0045] Furthermore, if the fire is small, based on the direction and distance, the tunnel inspection robot is controlled to carry fire-fighting equipment (such as fire extinguishers, fire bombs, etc.) to perform initial fire-fighting. If the fire is large, the tunnel inspection robot is controlled to evacuate the vehicles in the tunnel.

[0046] Step 104: Based on the opening operation and the closing operation, the crack features of the tunnel image are magnified and extracted.

[0047] In one embodiment of the present application, the crack feature extraction in the tunnel adopts a morphological algorithm. The single erosion and expansion operation cannot connect the crack breaks while removing noise. Erosion will reduce or even eliminate the crack area, and expansion will make the small noise larger or even merge with the crack. Therefore, it is necessary to combine the two operations to form an opening and closing operation. The opening operation is a process of first corroding and then expanding. It can eliminate the protruding burrs on the crack and make the overall contour smooth. At the same time, it minimizes the change in the area of ​​the crack connected domain and effectively eliminates isolated small noise in the image. The closing operation is a process of first expanding and then corroding. It can fill the holes in the crack area in the image and make the connected domain complete and smooth.

[0048] Step 105: based on the pixel grayscale values ​​corresponding to the tunnel image, the tunnel image after crack feature amplification is divided into a plurality of different connected domains to perform noise filtering on the tunnel image, and crack detection is performed on the highway tunnel based on the noise filtered tunnel image.

[0049] In one embodiment of the present application, based on the number of correct boxes of target detection results corresponding to the tunnel sample image category and the total number of all boxes detected for the tunnel sample image category, the prediction box accuracy ratio corresponding to the tunnel sample image category is determined. Based on the prediction box accuracy ratio and the number of images corresponding to the tunnel sample image category, the average accuracy rate corresponding to the tunnel sample image category is determined. According to the number of image categories and the average accuracy rate corresponding to each category of images, the average accuracy rate corresponding to the preset crack detection model is determined.

[0050] Specifically, in the target detection process, the preset crack detection model outputs several prediction boxes after calculating the input image, and the corresponding evaluation index is calculated based on the prediction box to determine whether it is correct. Specifically, the accuracy of the i-th sample Ii and the j-th category target Cj in the data set indicates the proportion of the target Cj detected by the prediction box in the category, that is, the ratio of the number of correctly predicted boxes containing the target Cj in the image to all predicted boxes containing the j-class target: Among them, TP is a true positive example, FP is a false positive example; AP is the average precision. When the validation set contains N images, the AP value is the average precision of the target category Cj in these images: Where N is the number of samples; and then the average precision of all target categories is the mean Where mAP is the mean of the average precision of all target categories in the set, and M is the number of target categories. Because tunnel lining cracks and water seepage can cause some accidents in serious cases, we are more concerned about whether cracks and water seepage can be fully detected.

[0051] In one embodiment of the present application, a crack feature enlarged image is obtained based on the tunnel image after the opening operation and the closing operation. The crack feature enlarged image is divided into a plurality of equal-sized regions. Based on the pixel grayscale values ​​corresponding to the plurality of equal-sized regions, the crack feature enlarged image is divided into a plurality of connected regions. The areas corresponding to the plurality of connected regions are determined, and the crack feature image is denoised based on the areas corresponding to the plurality of connected regions.

[0052] Specifically, first, the crack feature enlarged image is divided into multiple regions of equal size, namely, image blocks, in order to facilitate the subsequent separate processing of each small region, thereby improving processing efficiency and reducing computational complexity. By distinguishing cracks from background or other noise, multiple independent connected regions are formed. The number of pixels in each connected region is counted, and the area of ​​each connected region is calculated to perform denoising on the crack feature image.

[0053] In one embodiment of the present application, when the pixel grayscale values ​​of the four points on the left, upper left, top and upper right in any equal-sized area are all 0, they are marked as the starting point of a new connected domain. Alternatively, when the pixel grayscale values ​​of the two points on the left and upper left in any equal-sized area are all 1, the smaller point of the two pixel marking values ​​on the left and upper left is used as the marking point. Alternatively, when the pixel grayscale values ​​of the two points on the upper left and upper right in any equal-sized area are all 1, the smaller point of the two pixels on the upper left and upper right is used as the marking point. Alternatively, the marking points are marked in the order of left, upper left, top and upper right. Traverse multiple equal-sized areas, and record the areas with the same mark as a connected domain.

[0054] Specifically, for each equal-sized region, a marking system is initialized to record the starting point and boundary of the connected region. Each pixel can have a marking value associated with it, which can be undefined or set to a specific initial value such as 0 at the beginning. In each equal-sized region, the grayscale value of the boundary pixel is checked to determine whether a new connected region is started. For example, if the grayscale values ​​of the left and upper left pixels are both 0, the current pixel can be marked as the starting point of a new connected region. If the grayscale values ​​of the upper left and upper right pixels are both 1, the point with the smaller grayscale value is selected as the marking point. This helps to merge adjacent connected regions in subsequent processing.

[0055] Furthermore, the pixels in each equal-sized region are traversed from left to right and from top to bottom. If the grayscale values ​​of the left and upper-left pixels are both 0, the current pixel is marked as the starting point of a new connected region. If one of the grayscale values ​​of the left and upper-left pixels is 1, and the other is not marked or the marked value is greater than the grayscale value of the current pixel, the mark value of the pixel is updated to the grayscale value of the current pixel. If the grayscale values ​​of the upper and upper-right pixels are both 1, and at least one of them is not marked or the marked value is greater than the grayscale value of the current pixel, the point with the smaller grayscale value is selected and its mark value is updated to the grayscale value of the current pixel. During the traversal process, if the grayscale value of the current pixel is the same as the pixel on its left or above, and these pixels have been marked, the current pixel should inherit the same mark value.

[0056] Furthermore, during the traversal process, it is found that some connected regions with adjacent label values ​​should actually be the same connected region. Therefore, after the traversal, it is necessary to check all label values ​​and merge those connected regions with similar or adjacent grayscale values. Repeat the above steps to traverse all equal-sized regions in the image until all regions have been processed.

[0057] Further, the areas corresponding to the multiple connected regions are determined, and the areas of the minimum circumscribed rectangles corresponding to the multiple connected regions are determined. When the ratio between the area corresponding to the current connected region and the area of ​​the minimum circumscribed rectangle corresponding to the current connected region is greater than the preset connected domain pixel ratio, the current connected region is retained, otherwise the current connected region is used as the background to perform denoising on the crack feature image.

[0058] Furthermore, for each marked connected region, all its pixels are traversed and counted to obtain the area of ​​the connected region. Area calculation can be simply achieved by accumulating the number of all pixels in the connected region. For each connected region, all pixels in the connected region are traversed, their minimum and maximum coordinates on the X-axis and Y-axis are found, the positions of the upper left corner and the lower right corner of the rectangle are determined according to these coordinates, and the width and height of the rectangle are calculated to obtain the area of ​​the minimum circumscribed rectangle. For each connected region, the ratio between its area and the area of ​​the corresponding minimum circumscribed rectangle is calculated. This ratio can determine the compactness of the connected region, that is, the proportion of the connected region to its circumscribed rectangle. Based on the calculated area ratio, it is compared with the preset connected domain pixel ratio threshold. If the area ratio of the current connected region is greater than the threshold, it is considered to be a valid crack region and is retained; otherwise, it is regarded as background noise and removed.

[0059] Furthermore, for all marked connected domains, the area S of the connected domain and the minimum bounding rectangle of the connected domain are calculated respectively. The length and width of the minimum bounding rectangle are recorded as A and B, satisfying And S>TS is retained, and the connected domain that does not meet the requirement is recorded as the background. T is the preset connected domain pixel ratio threshold. After the above steps are completed, the information characteristics of the crack are preserved to a large extent.

[0060] In one embodiment of the present application, based on the feature-extracted image corresponding to the current highway tunnel, the area of ​​the crack in the image is obtained. Based on the area of ​​the crack in the image, the actual area of ​​the highway tunnel corresponding to the image, and the ratio between the image and the actual area, the actual area of ​​the crack in the tunnel is determined.

[0061] Step 106: Based on the fire incidence rate and crack conditions, determine the potential safety hazard of the highway tunnel.

[0062] In one embodiment of the present application, when the fire incidence rate is greater than a preset incidence rate threshold, or the crack ratio is greater than a preset crack ratio threshold, it is determined that there is a safety hazard in the current highway tunnel, and an alarm is issued to remind the patrol personnel of the current highway tunnel.

[0063] Figure 2A schematic diagram of a road tunnel safety hazard detection device provided in an embodiment of the present application. Figure 2 As shown, a road tunnel safety hazard detection device 200 includes: at least one processor 201; and a memory 202 in communication with the at least one processor 201; wherein the memory 202 stores instructions that can be executed by the at least one processor 201, and the instructions are executed by the at least one processor 201, so that the at least one processor 201 can: obtain environmental information and tunnel images in the road tunnel through multiple sensor components set at different detection points in the road tunnel; determine the correlation between the different detection points based on the environmental information, and based on the correlation, the temperature information corresponding to the different detection points and the preset temperature prediction model, The fire incidence rate in the highway tunnel is detected; when the detection result is a fire, the fire severity level is predicted based on the fire information obtained by the sensor, and the fire is processed based on the prediction result; based on opening and closing operations, the crack features of the tunnel image are amplified and extracted; based on the pixel grayscale values ​​corresponding to the tunnel image, the tunnel image after the crack features are amplified is divided into multiple different connected domains to filter the noise of the tunnel image, and the crack situation of the highway tunnel is detected based on the tunnel image after the noise is filtered; based on the fire incidence rate and the crack situation, the safety hazard situation of the highway tunnel is determined.

[0064] The embodiment of the present application also provides a non-volatile computer storage medium storing computer executable instructions, wherein the computer executable instructions are configured to: obtain environmental information and a tunnel image in a highway tunnel by means of a plurality of sensor components disposed at different detection points in the highway tunnel; determine the correlation between the different detection points based on the environmental information, and detect the fire incidence rate in the highway tunnel based on the correlation, the temperature information corresponding to the different detection points, and a preset temperature prediction model; when the detection result is that a fire has occurred, predict the severity level of the fire based on the fire information obtained by the sensor, and process the fire based on the prediction result; amplify and extract crack features from the tunnel image based on an open operation and a closed operation; divide the tunnel image after crack feature amplification into a plurality of different connected domains based on the pixel grayscale values ​​corresponding to the tunnel image, so as to perform noise filtering on the tunnel image, and detect crack conditions in the highway tunnel based on the tunnel image after noise filtering; determine the safety hazard of the highway tunnel based on the fire incidence rate and the crack conditions.

[0065] Each embodiment in this application is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device, equipment, and non-volatile computer storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0066] The above describes specific embodiments of the present application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0067] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the embodiments of the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the embodiments of the present application should be included in the scope of the claims of the present application.

Claims

1. A method for detecting potential safety hazards in a highway tunnel, characterized in that: The method comprises: The environmental information and tunnel image in the highway tunnel are obtained by setting up multiple sensor components at different detection points in the highway tunnel; Determine the correlation between the different detection points based on the environmental information, and detect the fire incidence in the highway tunnel based on the correlation, the temperature information corresponding to the different detection points, and a preset temperature prediction model; When the detection result indicates that a fire has occurred, the severity of the fire is predicted based on the fire information obtained by the sensor, and the fire is handled based on the prediction result; Based on opening and closing operations, amplifying and extracting crack features of the tunnel image; Based on the pixel grayscale values ​​corresponding to the tunnel image, the tunnel image after crack feature amplification is divided into a plurality of different connected domains to perform noise filtering on the tunnel image, and crack detection is performed on the highway tunnel based on the tunnel image after noise filtering; Based on the fire incidence rate and the crack conditions, determining the potential safety hazard of the highway tunnel; The determining the correlation between the different detection points based on the environmental information specifically includes: Determine the Euclidean distance between the different detection points under different environmental parameters; wherein the environmental parameters include at least one of temperature information, flame information and smoke information; Establishing a correlation matrix based on the Euclidean distance, and determining reference correlations between different detection points based on the correlation matrix; Acquire wind direction information and wind speed information in the tunnel in real time, and determine a first reference weight value in a first preset weight database based on the wind direction information and the wind speed information; Acquire tunnel layout information between the different detection points, and determine a second reference weight value in a second preset weight database based on the tunnel layout information; Based on the reference correlation, the first reference weight value and the second reference weight value, the correlation between the different detection points is determined.

2. A method for detecting potential safety hazards in highway tunnels according to claim 1, characterized in that: The predicting of the fire incidence rate in the highway tunnel based on the correlation degree, the temperature information corresponding to different detection points and the preset temperature prediction model specifically includes: The temperature information obtained by multiple sensors is input into a preset temperature prediction model to determine the predicted temperature information corresponding to multiple detection points in the tunnel at multiple time points in the future; Comparing the predicted temperature information corresponding to the multiple future time points with the preset reference temperature information to determine the temperature difference sets corresponding to the multiple detection points; wherein the preset reference temperature information includes the critical temperatures for fire occurrence corresponding to different times in the tunnel; Based on the correlation, determining the temperature influence values ​​of the multiple detection points on the current detection point; wherein the temperature influence value indicates the influence degree of the temperature change of one detection point on the temperature of another detection point; According to the temperature impact value and the temperature difference value sets respectively corresponding to the plurality of detection points, the predicted temperature information corresponding to the current detection point is adjusted to determine the fire occurrence rate of the current detection point based on the adjusted predicted temperature information.

3. A method for detecting potential safety hazards in highway tunnels according to claim 1, characterized in that: The method of predicting the severity of a fire based on the fire information acquired by the sensor and processing the fire based on the prediction result specifically includes: Obtaining flame information respectively sent by a plurality of flame sensors within a preset time period, and determining the location of the fire in the highway tunnel based on the flame information; Acquire temperature information respectively sent by a plurality of temperature sensors within a preset time period, and determine the distance between the fire location and the tunnel inspection robot based on the temperature information; Acquire multiple smoke information sent by the smoke sensor within a preset time period, so as to detect the current fire severity level based on the smoke information, the flame information and the temperature information; Based on the detection results, different processing instructions are issued to the tunnel inspection robot, and based on the orientation and the distance, the tunnel inspection robot is controlled to perform fire processing.

4. A method for detecting potential safety hazards in highway tunnels according to claim 1, characterized in that: The step of dividing the tunnel image after crack feature amplification into a plurality of different connected domains based on the pixel grayscale values ​​corresponding to the tunnel image, so as to perform noise filtering on the tunnel image, specifically includes: Based on the tunnel image after the opening operation and the closing operation, an amplified image of the crack characteristics is obtained; Dividing the crack feature magnified image into a plurality of regions of equal size; Dividing the crack feature magnified image into a plurality of connected regions based on the pixel grayscale values ​​respectively corresponding to the plurality of regions of equal size; The areas respectively corresponding to the plurality of connected regions are determined, so as to perform denoising processing on the crack feature image based on the areas respectively corresponding to the plurality of connected regions.

5. A method for detecting safety hazards in highway tunnels according to claim 4, characterized in that: The method of dividing the crack feature magnified image into a plurality of connected regions based on the pixel grayscale values ​​respectively corresponding to the plurality of regions of equal size specifically includes: In any of the equal-sized regions, when the pixel grayscale values ​​of the four points on the left, upper left, upper and upper right are all 0, mark it as the starting point of a new connected domain; Alternatively, in the case where the pixel grayscale values ​​of the left and upper left points in any of the equal-sized regions are both 1, the point with the smaller pixel mark value between the left and upper left points is used as the mark point; Alternatively, in any of the equal-sized regions, the pixel grayscale values ​​of the upper left and upper right points are both 1, and the smaller point of the upper left and upper right pixels is used as the marking point; Alternatively, the marking points are marked in the order of left side, upper left, upper side, and upper right; The multiple regions of equal size are traversed, and the regions with the same mark are recorded as a connected region.

6. A method for detecting potential safety hazards in highway tunnels according to claim 4, characterized in that: The determining of the areas respectively corresponding to the plurality of connected regions, and performing denoising processing on the crack feature image based on the areas respectively corresponding to the plurality of connected regions, specifically includes: Determine the areas corresponding to the multiple connected regions, and determine the areas of the minimum circumscribed rectangles corresponding to the multiple connected regions; When the ratio of the area corresponding to the current connected region to the area of ​​the minimum circumscribed rectangle corresponding to the current connected region is greater than the preset connected domain pixel ratio, the current connected region is retained; otherwise, the current connected region is used as the background to denoise the crack feature image.

7. A method for detecting potential safety hazards in a highway tunnel according to claim 1, characterized in that: Before performing crack detection on the highway tunnel based on the tunnel image after noise filtering, the method further includes: Based on the number of correct boxes of the target detection results corresponding to the tunnel sample image category and the total number of all boxes detected for the tunnel sample image category, determining the accuracy ratio of the prediction box corresponding to the tunnel sample image category; Based on the prediction box accuracy ratio and the number of images corresponding to the tunnel sample image category, determining an average accuracy value corresponding to the tunnel sample image category; According to the number of image categories and the average accuracy rate corresponding to each category of images, the average accuracy rate corresponding to the preset crack detection model is determined.

8. A road tunnel safety hazard detection device, characterized in that: The device comprises a memory for storing computer program instructions and a processor for executing the program instructions, wherein when the computer program instructions are executed by the processor, the device is triggered to execute the method according to any one of claims 1 to 7.

9. A non-volatile computer storage medium storing computer executable instructions, characterized in that: The computer executable instructions can execute the method according to any one of claims 1 to 7.

Images