Automatic identification and remote real-time monitoring method for water spraying amount of aproll of roadway roof

Abstract

The invention discloses an automatic identification and remote real-time monitoring method for aproll and water spraying amount of a roadway roof, which comprises the following steps: acquiring infrared thermal imaging data and visible light image data of the roadway roof by using an infrared thermal imaging camera and a visible light camera, and fusing the infrared thermal imaging data and the visible light image data into multispectral image data; segmenting a water dispersing or spraying area through an image processing algorithm, and calculating a water dispersing area; calculating the temperature difference between the aproll area and the background area based on the infrared thermal imaging data; extracting water dripping frequency and speed change to obtain dynamic characteristics of a water dripping track; and carrying out fusion analysis on the apron area, the temperature difference and the dynamic characteristic data to generate a water damage intensity index, and carrying out future water damage intensity prediction by using the water damage intensity index in a period of time. According to the method, the identification precision of the aproll area of the roadway roof and the real-time performance of water disaster monitoring are remarkably improved, and meanwhile, the potential risk caused by missing detection or false detection is reduced, so that more reliable technical support is provided for mine safety management.

Description

Technical Field

[0001] The present invention relates to the technical field of roadway roof safety monitoring, and particularly relates to a method for automatically identifying the water spraying and dripping volume on the roadway roof and remotely real-time monitoring. Background Art

[0002] The phenomenon of water spraying or dripping on the mine roof is one of the common water hazard problems in the mine production process, which seriously affects the safety and production efficiency of mine operations. If the phenomenon of water spraying or dripping on the roof is not discovered and handled in time, it may lead to roof collapse, equipment damage and personal safety accidents. Therefore, establishing an efficient and accurate roadway roof water hazard monitoring and early warning system is crucial for mine safety management.

[0003] The existing mine roof water hazard monitoring methods mainly include the drilling monitoring method and the hydrogeological observation method. The drilling monitoring method analyzes the aquifers and leakage conditions in the mine strata by arranging boreholes and combining water level measurement and flow observation; the hydrogeological observation method relies on systematically monitoring the distribution and flow characteristics of surface water and groundwater in the mining area to judge the possible areas and intensities of water hazards. These methods provide basic data support for the prevention of mine water hazards, but usually require complex equipment and a large amount of manual participation, and the monitoring process is time-consuming and it is difficult to achieve real-time and dynamic monitoring.

[0004] Although the drilling monitoring method has high accuracy, the borehole layout process is complex and costly, and it is difficult to arrange in a large area; the hydrogeological observation method relies on the long-term accumulation of hydrogeological data and cannot quickly reflect the dynamic changes of the water spraying or dripping phenomenon. These traditional methods lack intelligent analysis means and cannot predict the development trend of water hazards through multi-source data, and the early warning ability is limited. To solve these problems, there is an urgent need for a comprehensive solution based on multi-spectral vision technology, deep learning algorithms and remote real-time monitoring to achieve high-precision, dynamic and intelligent management of water hazard monitoring. Summary of the Invention

[0005] The present invention adopts the following technical solutions:

[0006] A method for automatically identifying the water spraying and dripping volume on the roadway roof and remotely real-time monitoring, the method comprising the following steps:

[0007] Step 1: Use an infrared thermal imaging camera and a visible light camera to obtain infrared thermal imaging data and visible light image data of the roadway roof, and fuse the infrared thermal imaging data and the visible light image data into multi-spectral image data;

[0008] Step 2: Segment the water spraying or dripping area through an image processing algorithm, and calculate the water spraying area A ir ;

[0009] Step 3: Calculate the temperature difference ΔT between the water-sprinkling area and the background area based on the infrared thermal imaging data;

[0010] Step 4: Extract the dripping frequency and speed changes to obtain the dynamic feature R of the dripping trajectory v ;

[0011] Step 5: Fuse and analyze the water-sprinkling area, temperature difference, and dynamic feature data to generate the water damage intensity index I s , and its calculation formula is: I s = w1·A ir + w2·ΔT + w3·R v ; where w1, w2, and w3 are weighting coefficients; according to the obtained water damage intensity index I s , trigger multi-level alarms;

[0012] Step 6: Use the water damage intensity index I over a period of time s to predict the future water damage intensity.

[0013] Preferably, the infrared thermal imaging data is used to record the temperature distribution of the roadway roof, and the visible light image is used to record the dripping trajectory and the boundary of the water-sprinkling area on the surface of the roadway roof. Align the time stamps of the infrared thermal imaging data and the visible light image data to generate unified time series data D(t), which is the multi-spectral image data;

[0014] D(t) = {D IR (t), D RGB (t)};

[0015] where D IR (t) is the infrared thermal imaging data at time t, and D RGB (t) is the visible light image data at time t.

[0016] Preferably, Step 2 specifically includes:

[0017] Step 2.1: Denoise the multi-spectral image data, and use the Gaussian filtering algorithm to reduce random noise interference; optimize the boundary of the water-sprinkling area through erosion and dilation processing to eliminate isolated noise points; convert the multi-spectral image data into a grayscale image;

[0018] Step 2.2: Segment the water-sprinkling area, and use the Mask R-CNN algorithm to segment the grayscale image to generate a binary image. The pixels in the water-sprinkling area of the binary image are 1, and the pixels in other areas are 0;

[0019] Step 2.3: Statistically calculate the total area of the area with pixel value 1 to obtain the water-sprinkling area A ir .

[0020] Preferably, Step 3 specifically includes:

[0021] Calculate the temperature difference ΔT between the apron area and the background area using the infrared thermal imaging data:

[0022]

[0023] where B is the boundary of the apron area, T(x, y) is the pixel temperature value, N water is the pixel temperature value of the apron area, N background is the pixel temperature value of the background area.

[0024] Preferably, analyze the water droplet trajectories in the continuous multi-spectral image data using the inter-frame difference method, and extract the dripping frequency f drop and velocity v drop , and input the dripping frequency f drop and velocity v drop into the 3D-CNN model to generate the dynamic feature R v :

[0025] R v = w f ·f drop + w v ·v drop ;

[0026] w f and w v are the weight coefficients of the dripping frequency and the velocity respectively.

[0027] Preferably, triggering a multi-level alarm means that: when I s < T1, a first-level alarm is given; when T1 ≤ I s < T2, a second-level alarm is given; when I s ≥ T2, a third-level alarm is given.

[0028] Preferably, step 6 specifically includes:

[0029] Input the historical water disaster intensity index into the LSTM model to train and form a water disaster intensity index prediction model, and input the water disaster intensity index of the past k time points into the water disaster intensity index prediction model to obtain the predicted value of the water disaster intensity index at the next time point.

[0030] The beneficial effects of the present invention are:

[0031] By fusing infrared thermal imaging data and visible light image data into multi-spectral image data, this invention uses deep learning algorithms to analyze the image features of the water seepage area on the roadway roof, extracts the water seepage area, temperature difference, and dynamic water dripping features, and performs weighted fusion processing on these features to generate a water disaster intensity index, realizing the dynamic monitoring and multi-level alarm mechanism for the water seepage area of the roadway; on this basis, combined with the time series prediction model, it can judge the development trend of water disasters in advance. This method significantly improves the recognition accuracy of the water seepage area on the roadway roof and the real-time performance of water disaster monitoring, while reducing the potential risks caused by missed inspections or misjudgments, thus providing more reliable technical support for mine safety management. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 is a flowchart of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0033] The following further describes the specific implementation manners of the present invention in conjunction with the drawings and specific embodiments:

[0034] Combined with Figure 1 , an automatic recognition and remote real-time monitoring method for the water dripping volume on the roadway roof, the method includes the following steps:

[0035] Step 1: Use an infrared thermal imaging camera and a visible light camera to obtain infrared thermal imaging data and visible light image data of the roadway roof, and fuse the infrared thermal imaging data and visible light image data into multi-spectral image data.

[0036] The infrared thermal imaging data is used to record the temperature distribution of the roadway roof, and the visible light image is used to record the water dripping trajectory and the boundary of the water seepage area on the roadway roof surface. The collected infrared thermal imaging data and visible light image data are transmitted back to the edge computing device.

[0037] In the data collection stage, the resolution and temperature sensitivity of the infrared thermal imaging camera should meet the monitoring requirements in the complex mine environment, and the visible light camera should have the imaging ability under low light conditions to ensure that the image data of the water seepage area can be clearly recorded regardless of the light intensity in the roadway.

[0038] Align the time stamps of the infrared thermal imaging data and the visible light image data to generate unified time series data D(t), which is the multi-spectral image data;

[0039] D(t) = {D IR (t), D RGB (t)};

[0040] wherein, D IR (t) is the infrared thermal imaging data at time t, and D RGB (t) is the visible light image data at time t.

[0041] There may be interferences such as environmental noise and light reflection in the multi-spectral image data of the roadway roof. Therefore, it is necessary to preprocess the data to extract effective features.

[0042] Step 2: The edge computing device segments the water spraying or watering area through an image processing algorithm and calculates the water spraying area A ir 。

[0043] Specifically, it includes: Step 2.1: Denoise the multi-spectral image data, and use the Gaussian filtering algorithm to reduce the interference of random noise; optimize the boundary of the water spraying area through erosion and dilation processing to eliminate isolated noise points; convert the multi-spectral image data into a grayscale image.

[0044] Step 2.2: Segment the water spraying area, use the Mask R-CNN algorithm to segment the grayscale image to generate a binary image. The pixels in the water spraying area of the binary image are 1, and the pixels in other areas are 0.

[0045] Step 2.3: Statistically calculate the total area of the area with a pixel value of 1 to obtain the water spraying area A ir 。

[0046] The image segmentation algorithm adopts the Mask R-CNN model, and optimizes the training parameters in combination with the characteristics of the roadway environment to ensure the accuracy of the segmentation results. Through morphological operations, including erosion and dilation processing, further eliminate the noise interference in the image and optimize the boundary of the water spraying area.

[0047] Step 3: The edge computing device calculates the temperature difference ΔT between the water spraying area and the background area based on the infrared thermal imaging data.

[0048] Specifically, it includes:

[0049] Calculate the temperature difference ΔT between the water spraying area and the background area using the infrared thermal imaging data:

[0050]

[0051] where B is the boundary of the water spraying area, T(x, y) is the pixel temperature value, N water is the pixel temperature value of the water spraying area, N background is the pixel temperature value of the background area.

[0052] Step 4: The edge computing device extracts the dripping frequency and speed change to obtain the dynamic feature R of the dripping trajectory v 。

[0053] Analyze the water droplet trajectory in the continuous multi-spectral image data using the inter-frame difference method to extract the dripping frequency f drop and speed v drop, the dripping frequency f drop and the speed v drop are input into the 3D-CNN model to generate the dynamic feature R v :

[0054] R v = w f ·f drop + w v ·v drop ;

[0055] w f and w v are the weight coefficients of the dripping frequency and the speed respectively.

[0056] Step 5: The edge computing device transmits the watering area, temperature difference and dynamic feature data back to the cloud platform through the wireless network.

[0057] The cloud platform stores, analyzes and alarms the received data, and provides a dynamic visualization interface to display the location, intensity of the watering area and predict the change trend.

[0058] The cloud platform fuses and analyzes the watering area, temperature difference and dynamic feature data to generate the water disaster intensity index I s , and its calculation formula is: I s = w1·A ir + w2·ΔT + w3·R v ; where w1, w2, w3 are the weighting coefficients.

[0059] In the water disaster intensity assessment stage, the weight coefficients of the watering area, temperature difference and dynamic feature are calibrated through experiments according to different scenarios of the mine; for areas with a larger watering area, a higher weight is assigned; for areas with a smaller temperature difference but significant dripping dynamic features, the corresponding weights are dynamically adjusted to ensure the reliability of the assessment results.

[0060] The cloud platform triggers multi-level alarms according to the obtained water disaster intensity index I s .

[0061] Specifically: when I s < T1, a first-level alarm is issued; when T1 ≤ I s < T2, a second-level alarm is issued; when I s ≥ T2, a third-level alarm is issued.

[0062] The first-level alarm indicates low risk, and the staff conducts daily inspections. The second-level alarm indicates medium risk, and the staff needs to conduct on-site inspections. The third-level alarm indicates high risk, and an emergency response is initiated, and operations such as stopping work and draining water are required.

[0063] Step 6: The cloud platform uses the water disaster intensity index I s, to predict the future water disaster intensity.

[0064] Specifically, it includes: inputting historical water disaster intensity indicators into the LSTM model for training to form a water disaster intensity indicator prediction model, and inputting the water disaster intensity indicators at the past k time points into the water disaster intensity indicator prediction model to obtain the predicted value of the water disaster intensity indicator at the next time point.

[0065] The cloud monitoring platform provides a three-dimensional visualization interface to superimpose and display the real-time state and historical changes of the roadway roof, enhancing the intuitiveness of data interpretation.

[0066] Certainly, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions, or substitutions made by those skilled in the art within the essence of the present invention should also fall within the protection scope of the present invention.

Claims

1. An automatic recognition and remote real-time monitoring method for the water spray and water inflow on the roadway roof, characterized in that, The method includes the following steps: Step 1: Use an infrared thermal imaging camera and a visible light camera to obtain infrared thermal imaging data and visible light image data of the roadway roof, and fuse the infrared thermal imaging data and the visible light image data into multi-spectral image data; Step 2: Segment the watering or sprinkling area through an image processing algorithm and calculate the watering area A ir ; Step 3: Calculate the temperature difference ΔT between the water-dispersion area and the background area based on the infrared thermal imaging data; Step 4: Extract the dripping frequency and speed changes to obtain the dynamic feature R of the dripping trajectory v ; Step 5: Fuse and analyze the apron area, temperature difference, and dynamic characteristic data to generate the water damage intensity index I s , and its calculation formula is: I s = w1·A ir + w2·ΔT + w3·R v ; where w1, w2, and w3 are weighting coefficients; according to the obtained water damage intensity index I s , trigger multi-level alarms; Step 6: Use the water damage intensity index I for a period of time s to predict the future water damage intensity.

2. The automatic identification and remote real-time monitoring method for the water spraying volume of the roadway roof according to claim 1, characterized in that The infrared thermal imaging data is used to record the temperature distribution of the roadway roof, and the visible light image is used to record the dripping water trajectory and the boundary of the water-dispersion area on the surface of the roadway roof. Align the time stamps of the infrared thermal imaging data and the visible light image data to generate unified time series data D(t), which is the multi-spectral image data; D(t) = {D IR (t), D RGB (t)}; Among them, D IR (t) is the infrared thermal imaging data at time t, and D RGB (t) is the visible light image data at time t.

3. The automatic identification and remote real-time monitoring method for the water spray and water quantity of the roadway roof according to claim 1, characterized in that Step 2 specifically includes: Step 2.1: Denoise the multi-spectral image data, and use the Gaussian filtering algorithm to reduce random noise interference; Optimize the boundary of the water-dispersion area through erosion and dilation processing to eliminate isolated noise points; Convert the multi-spectral image data into a grayscale image; Step 2.2: Segment the water-dispersion area, and use the Mask R-CNN algorithm to segment the grayscale image to generate a binary image. The pixels in the water-dispersion area of the binary image are 1, and the pixels in other areas are 0; Step 2.3: Calculate the total area of the regions with pixel value 1 to obtain the area A of the apron region ir .

4. The automatic recognition and remote real-time monitoring method for the water spray amount on the roadway roof according to claim 1, characterized in that Step 3 specifically includes: Calculate the temperature difference ΔT between the water-dispersion area and the background area using the infrared thermal imaging data: Where B is the boundary of the apron area, T(x, y) is the pixel temperature value, N water is the pixel temperature value of the apron area, N background is the pixel temperature value of the background area.

5. The automatic identification and remote real-time monitoring method for the water spray amount on the roadway roof according to claim 1, characterized in that Analyze the water droplet trajectories in continuous multi-spectral image data using the inter-frame difference method, and extract the dripping frequency f drop and the velocity v drop , and input the dripping frequency f drop and the velocity v drop into the 3D-CNN model to generate the dynamic feature R v : R v = w f · f drop + w v · v drop ; w f and w v are the weight coefficients of the dripping frequency and speed, respectively.

6. The automatic recognition and remote real-time monitoring method for the water spray quantity of the roadway roof according to claim 1, wherein Triggering multi-level alarms means that when I s < T1, a first-level alarm is triggered; when T1 ≤ I s < T2, a second-level alarm is triggered; when I s ≥ T2, a third-level alarm is triggered.

7. A method for automatically identifying the water spray amount on the roadway roof and remotely real-time monitoring, according to claim 1, characterized in that, Step 6 specifically includes: Input the historical water disaster intensity index into the LSTM model for training to form a water disaster intensity index prediction model. Input the water disaster intensity index of the past k time points into the water disaster intensity index prediction model to obtain the predicted value of the water disaster intensity index at the next time point.

Images