A green construction method for subway track connection

Abstract

The present invention discloses a green construction method for a subway track connection body, which belongs to the field of track construction. The specific steps of the construction method are as follows: (1) collecting track material information and classifying and arranging it; (2) selecting and splicing track materials and performing error correction analysis; (3) collecting waste generated by construction and centrally processing it; (4) collecting construction information in real time and feeding it back to a management platform; the present invention effectively improves the accuracy of tracking position by constructing a motion model, ensures that the tracking target can be fully matched across cameras, improves the judgment accuracy of track connection, ensures the safety and stability of track connection, can ensure the safety of workers in handling construction waste, and can adjust the deficiencies in the treatment plan, improve the efficiency and comprehensiveness of waste treatment, and prevent environmental pollution caused by improper treatment.

Description

Technical Field

[0001] The present invention relates to the field of track construction, and in particular to a green construction method for a subway track connection body. Background Art

[0002] The subway is a high-speed, high-capacity, electric-powered rail transit system built in cities. Trains run on fully enclosed lines. Lines in the central city are basically located in underground tunnels, and lines outside the central city are generally located on viaducts or on the ground. The subway is a high-density, high-capacity, and exclusive underground and above-ground right-of-way system in urban areas. Since urban land in large cities is generally valuable, building railways underground can save ground space and allow the ground land to be used for other purposes. At the same time, since the railway is built underground, ground noise can be reduced, and the subway route does not overlap or intersect with other routes, so driving is less disturbed by traffic, which can save a lot of commuting time. Therefore, the subway has become one of the important commuting methods for people, and subway construction technology has become increasingly important.

[0003] A search revealed Chinese patent number CN112144322A, which discloses a green track construction method. While this invention allows for rapid assembly and disassembly of longitudinal steel connections, replacing the existing welding method for green construction, it suffers from low tracking accuracy, reducing the accuracy of track connection determination. Furthermore, existing green construction methods for the main body of subway track connections fail to address deficiencies in the treatment solution, reducing the efficiency and comprehensiveness of waste treatment and making it prone to environmental pollution due to improper waste disposal. Therefore, we propose a green construction method for the main body of subway track connections. Summary of the Invention

[0004] The purpose of the present invention is to solve the defects in the prior art and to propose a green construction method for connecting the main body of a subway track.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A green construction method for connecting a subway track body, the specific steps of the construction method are as follows:

[0007] (1) Collect and classify track material information;

[0008] (2) Select and splice track materials and perform error correction analysis;

[0009] (3) Collect construction waste and treat it centrally;

[0010] (4) Collect construction information in real time and feed it back to the management platform.

[0011] As a further solution of the present invention, the specific steps of classification and sorting in step (1) are as follows:

[0012] Step 1: The staff uploads the track material information of each group to the server, and then the server classifies the received track material information into rails, sleepers, connecting parts, trackbed, track reinforcement equipment and switches;

[0013] Step 2: After that, each type of track material is numbered, and at the same time, a corresponding material three-dimensional model is constructed according to the parameter information of each group of track materials, and then each group of material models is numbered according to the number information of each group of materials.

[0014] As a further solution of the present invention, the specific steps for selecting the track material in step (2) are as follows:

[0015] Step I: The staff inputs the type and number of the required rail material through the computer. The computer then sends a material collection instruction to the intelligent collection vehicle. The intelligent collection vehicle receives the collection instruction and moves to the corresponding material warehouse location.

[0016] Step II: After that, it communicates with the server and collects the storage locations of each group of materials. Then, based on the material type and number information selected by the staff, it moves to the corresponding track material location. The intelligent collection vehicle collects the track material with its own collection arm and places the collected material in the transport cabin.

[0017] Step III: After the material collection is completed, the intelligent collection vehicle moves to the relevant staff. When the staff removes each set of track materials, the intelligent collection vehicle returns to the specified position and waits for the next instruction.

[0018] As a further solution of the present invention, the specific steps of the error correction analysis in step (2) are as follows:

[0019] Step 1: The monitoring equipment collects image information of the workers connecting the track materials, and converts the image space into frequency space and vice versa through Fourier transform, and filters the high-frequency components in the image information converted to frequency space;

[0020] Step 2: After the computer collects the image information of each set collected by each monitoring device, it processes the video or image sequence frames with a fixed frame rate offline, calculates the interval time of the actual video frames, records the calculated interval time of the actual video frames, and establishes a motion model through Kalman filter theory. At the same time, the motion state of the track material is obtained in real time through the constructed motion model;

[0021] Step 3: Collect the motion state of the track material in the current video frame and construct a prediction equation to estimate the motion state of each track material in the next video frame. At the same time, obtain the position set, representation position and representation covariance matrix in the current video frame and calculate the detection result of the i-th monitoring device and the Mahalanobis distance matrix of the tracking target;

[0022] Step 4: Calculate the cosine distance between the detection result and the tracked target based on the appearance feature vector of the tracked target and the set of appearance feature vectors of the detection result. Filter the calculated cosine distance based on the threshold. Then, perform binary matching between the detection result and the tracked target using the Hungarian algorithm. After the matching is completed, the computer processes the video frame data of each video stream in parallel and performs target labeling, estimation of the motion state of the tracked target, matching association, and real-time tracking of multiple targets across cameras on the downsampled video frames of each video stream.

[0023] Step 5: Real-time collection of noise decibels when workers are splicing track materials. When the noise decibel exceeds the threshold set by the workers, the computer issues a warning and reduces the operating power of the equipment. At the same time, based on the matching results, the corresponding three-dimensional model of the material is extracted from the server for simulation. At the same time, it detects whether there is position offset or loose connection in the splicing. If so, a warning is issued to the workers, prompting them to make corrections.

[0024] As a further solution of the present invention, the specific transformation formula of the Fourier forward and inverse transform in the first step is as follows:

[0025]

[0026]

[0027] Where u and v are both frequency variables, x and y are the coordinates of each pixel point of the image information, formula (1) is the Fourier forward transform, and formula (2) is the Fourier inverse transform;

[0028] The specific calculation formula for the interval time in the second step is as follows:

[0029]

[0030]

[0031] Where Δt k+1 Represents the interval time between two sets of video frames, Represents the delay time between the downsampled video frame and the original video stream, Represents the time it takes for the tracking algorithm to process a video frame.

[0032] As a further solution of the present invention, the specific steps of the centralized waste treatment in step (4) are as follows:

[0033] S1: Collect waste information generated during the construction process and classify it into hazardous waste and non-hazardous waste. Record the actual type and quantity of each waste group, and then transport the hazardous waste and non-hazardous waste to the corresponding areas;

[0034] S2: Receive the treatment plans uploaded by the staff, extract the corresponding treatment plans from the treatment database, then build a learning network model, and then import the treatment plans extracted from the database as sample data into the learning network model;

[0035] S3: The learning network model extracts waste treatment information from the sample data as feature data, normalizes each set of feature data, and then uses feature dimensionality reduction to select feature data that can represent treatment efficiency, while filtering out feature data with poor representation ability, and dividing them into training samples and test samples;

[0036] S4: The learning network model is trained through input, convolution, pooling, full connection, and output of the training samples. The trained learning network model is then tested using the test samples. The processing solution uploaded by the staff is then imported into the learning network model, and the relevant processing efficiency curve is output and analyzed.

[0037] S5: Adjust the treatment plan uploaded by the staff based on the analysis results, and feed the adjusted plan back to the staff for review. After the staff confirms that it is correct, they will treat the waste according to the treatment plan.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] 1. Compared with previous construction methods, the green construction method for the main body of subway track connection collects image information when workers connect track materials through monitoring equipment, optimizes the image information, and then processes video or image sequence frames with a fixed frame rate offline, calculates the interval time of actual video frames, records the calculated interval time of actual video frames, and establishes a motion model through Kalman filtering theory. At the same time, the motion state of the track material is obtained in real time through the constructed motion model, and then the detection result of the i-th monitoring device and the Mahalanobis distance matrix and cosine distance of the tracking target are calculated. The detection result and the tracking target are binary matched using the Hungarian algorithm. After the matching is completed, the computer processes the video frame data of each video stream in parallel, and performs target labeling, estimation of the motion state of the tracking target, matching association, and real-time tracking of multiple targets across cameras on the video frames obtained after downsampling in each video stream. By constructing the motion model, the accuracy of the tracking position is effectively improved, ensuring that the tracking target can be fully matched across cameras, improving the judgment accuracy of the track connection, and ensuring the safety and stability of the track connection.

[0040] 2. The green construction method of the subway track connection body collects and classifies waste information generated during the construction process through computers, and then builds a learning network model. The treatment plan extracted from the database is imported into the learning network model as sample data, and the waste treatment information is extracted from the sample data as feature data, and is divided into training samples and test samples. The training samples are used to train the learning network model through input, convolution, pooling, full connection and output, and the trained learning network model is tested through test samples. The treatment plan uploaded by the staff is then imported into the learning network model, and the relevant treatment efficiency curve is output and analyzed. The treatment plan uploaded by the staff is adjusted based on the analysis results, which can ensure the safety of the staff in handling construction waste, and at the same time, it can adjust the deficiencies in the treatment plan, improve the efficiency and comprehensiveness of waste treatment, and prevent environmental pollution due to improper treatment. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0042] Figure 1 This is a flowchart of a green construction method for subway track connection body proposed by the present invention. DETAILED DESCRIPTION

[0043] Example 1

[0044] Reference Figure 1 A green construction method for connecting the main body of a subway track is provided. The specific steps of the construction method are as follows:

[0045] Collect track material information and classify it.

[0046] Specifically, the staff will upload the information of each group of track materials to the server, and then the server will classify the received track material information according to rails, sleepers, connecting parts, roadbed, track reinforcement equipment and switches, and then number each type of track material. At the same time, according to the parameter information of each group of track materials, the corresponding three-dimensional material model will be constructed, and then the material model of each group will be numbered according to the number information of each group of materials.

[0047] Select and splice track materials and perform error correction analysis.

[0048] Specifically, the staff inputs the type and number of the track material to be used through the computer, and then the computer sends a material collection instruction to the intelligent collection vehicle. The intelligent collection vehicle receives the collection instruction and moves to the corresponding material warehouse location, then communicates with the server and collects the storage location of each group of materials. Then, according to the material type and number information selected by the staff, it moves to the corresponding track material. The track material is collected by the collection arm of the intelligent collection vehicle and the collected materials are placed in the transport cabin. After the material collection is completed, the intelligent collection vehicle moves to the relevant staff. When the staff removes each group of track materials, the intelligent collection vehicle returns to the specified position and waits for the next instruction.

[0049] Specifically, the monitoring equipment collects image information when the staff connects the track material, and performs mutual conversion from image space to frequency space through Fourier forward and inverse transform, and filters the high-frequency components in the image information converted to the frequency space. After the computer collects each set of image information collected by each monitoring device, it processes the video or image sequence frames with a fixed frame rate offline, and calculates the interval time of the actual video frames, records the calculated interval time of the actual video frames, and establishes a motion model through Kalman filtering theory. At the same time, the motion state of the track material is obtained in real time through the constructed motion model, and then the motion state of the track material in the current video frame is collected, and a prediction equation is constructed to estimate the motion state of each track material in the next video frame. At the same time, the position set, representation position and representation covariance matrix in the current video frame are obtained, and the detection result of the i-th monitoring device and the Mahalanobis distance matrix of the tracking target are calculated. According to the tracking target The cosine distance between the detection result and the tracking target is calculated based on the appearance feature vector of the target and the appearance feature vector set of the detection result, and the calculated cosine distance is filtered according to the threshold. The detection result and the tracking target are then binary matched by the Hungarian algorithm. After the matching is completed, the computer processes the video frame data of each video stream in parallel, and performs target marking, estimation of the motion state of the tracking target, matching association and real-time tracking of multiple targets across cameras on the video frames obtained after downsampling in each video stream. The noise collector collects the noise decibels when the staff splices the track materials in real time. When the noise decibel exceeds the threshold set by the staff, the computer issues a warning and reduces the operating power of the equipment. At the same time, the corresponding material three-dimensional model is extracted from the server according to the matching result for simulation. At the same time, it detects whether there is position offset and loose connection in the splicing. If so, a warning is issued to the staff and the staff is prompted to make corrections.

[0050] It should be further explained that the specific transformation formulas of Fourier forward and inverse transform are as follows:

[0051]

[0052]

[0053] Where u and v are both frequency variables, x and y are the coordinates of each pixel point of the image information, formula (1) is the Fourier forward transform, and formula (2) is the Fourier inverse transform;

[0054] The specific calculation formula for the interval time is as follows:

[0055]

[0056]

[0057] Where Δt k+1Represents the interval time between two sets of video frames, Represents the delay time between the downsampled video frame and the original video stream, Represents the time it takes for the tracking algorithm to process a video frame.

[0058] Example 2

[0059] Reference Figure 1 A green construction method for connecting the main body of a subway track is provided. The specific steps of the construction method are as follows:

[0060] Collect waste generated by construction and treat it centrally.

[0061] Specifically, the computer collects waste information generated during the construction process and classifies it into hazardous waste and non-hazardous waste, while recording the actual category and corresponding quantity information of each group of waste. Then, the hazardous waste and non-hazardous waste are transported to the corresponding areas respectively, and the treatment plan uploaded by the staff is received and the corresponding treatment plan is extracted from the treatment database. Then, a learning network model is constructed, and the treatment plan extracted from the database is imported into the learning network model as sample data. The learning network model extracts waste treatment information from the sample data as feature data and normalizes each group of feature data. Feature dimensionality reduction is then used to filter out feature data that can represent treatment efficiency, and feature data with poor representation ability is filtered out. The data is divided into training samples and test samples. The training samples are used to train the learning network model through input, convolution, pooling, full connection and output. The trained learning network model is then tested with test samples. The treatment plan uploaded by the staff is then imported into the learning network model, and the relevant treatment efficiency curve is output and analyzed. The treatment plan uploaded by the staff is adjusted based on the analysis results, and the adjusted plan is fed back to the staff for review. After the staff confirms that it is correct, they treat the waste according to the treatment plan.

[0062] Collect construction information in real time and feed it back to the management platform.

Claims

1. A green construction method for connecting the main body of a subway track, characterized in that: The specific steps of this construction method are as follows: (1) Collect and classify track material information; (2) Select and splice track materials and perform error correction analysis; (3) Collect construction waste and treat it centrally; (4) Collect construction information in real time and feed it back to the management platform; The specific steps of the error correction analysis in step (2) are as follows: Step 1: The monitoring equipment collects image information of the workers connecting the track materials, and converts the image space into frequency space and vice versa through Fourier transform, and filters the high-frequency components in the image information converted to frequency space; Step 2: After the computer collects the image information of each set collected by each monitoring device, it processes the video or image sequence frames with a fixed frame rate offline, calculates the interval time of the actual video frames, records the calculated interval time of the actual video frames, and establishes a motion model through Kalman filter theory. At the same time, the motion state of the track material is obtained in real time through the constructed motion model; Step 3: Collect the motion state of the track material in the current video frame and construct a prediction equation to estimate the motion state of each track material in the next video frame. At the same time, obtain the position set, representation position and representation covariance matrix in the current video frame and calculate the detection result of the i-th monitoring device and the Mahalanobis distance matrix of the tracking target; Step 4: Calculate the cosine distance between the detection result and the tracked target based on the appearance feature vector of the tracked target and the set of appearance feature vectors of the detection result. Filter the calculated cosine distance based on the threshold. Then, perform binary matching between the detection result and the tracked target using the Hungarian algorithm. After the matching is completed, the computer processes the video frame data of each video stream in parallel and performs target labeling, estimation of the motion state of the tracked target, matching association, and real-time tracking of multiple targets across cameras on the downsampled video frames of each video stream. Step 5: The noise decibel level is collected in real time when workers are splicing track materials. When the noise decibel level exceeds the threshold set by the workers, the computer issues a warning and reduces the equipment's operating power. Simultaneously, based on the matching results, the computer extracts the corresponding 3D model of the material from the server for simulation. It also detects whether there are any positional offsets or loose connections in the splicing. If so, the computer issues a warning and prompts the workers to make corrections. The specific transformation formula of the Fourier forward and inverse transform described in the first step is as follows: Where u and v are both frequency variables, x and y are the coordinates of each pixel point of the image information, formula (1) is the Fourier forward transform, and formula (2) is the Fourier inverse transform; The specific calculation formula for the interval time in the second step is as follows: Where Δt k+1 Represents the interval time between two sets of video frames, Represents the delay time between the downsampled video frame and the original video stream, Represents the time it takes for the tracking algorithm to process a video frame.

2. A green construction method for connecting a subway track according to claim 1, characterized in that: The specific steps of classification and sorting described in step (1) are as follows: Step 1: The staff uploads the track material information of each group to the server, and then the server classifies the received track material information into rails, sleepers, connecting parts, trackbed, track reinforcement equipment and switches; Step 2: After that, each type of track material is numbered, and at the same time, a corresponding material three-dimensional model is constructed according to the parameter information of each group of track materials, and then each group of material models is numbered according to the number information of each group of materials.

3. The green construction method for subway track connection body according to claim 1, characterized in that: The specific steps for selecting the track material in step (2) are as follows: Step I: The staff inputs the type and number of the required rail material through the computer. The computer then sends a material collection instruction to the intelligent collection vehicle. The intelligent collection vehicle receives the collection instruction and moves to the corresponding material warehouse location. Step II: After that, it communicates with the server and collects the storage locations of each group of materials. Then, based on the material type and number information selected by the staff, it moves to the corresponding track material location. The intelligent collection vehicle collects the track material with its own collection arm and places the collected material in the transport cabin. Step III: After the material collection is completed, the intelligent collection vehicle moves to the relevant staff. When the staff removes each set of track materials, the intelligent collection vehicle returns to the specified position and waits for the next instruction.

4. A green construction method for connecting a subway track according to claim 1, characterized in that: The specific steps for centralized waste treatment in step (4) are as follows: S1: Collect waste information generated during the construction process and classify it into hazardous waste and non-hazardous waste. Record the actual type and quantity of each waste group, and then transport the hazardous waste and non-hazardous waste to the corresponding areas; S2: Receive the treatment plans uploaded by the staff, extract the corresponding treatment plans from the treatment database, then build a learning network model, and then import the treatment plans extracted from the database as sample data into the learning network model; S3: The learning network model extracts waste treatment information from the sample data as feature data, normalizes each set of feature data, and then uses feature dimensionality reduction to select feature data that can represent treatment efficiency, while filtering out feature data with poor representation ability, and dividing them into training samples and test samples; S4: The learning network model is trained through input, convolution, pooling, full connection, and output of the training samples. The trained learning network model is then tested using the test samples. The processing solution uploaded by the staff is then imported into the learning network model, and the relevant processing efficiency curve is output and analyzed. S5: Adjust the treatment plan uploaded by the staff based on the analysis results, and feed the adjusted plan back to the staff for review. After the staff confirms that it is correct, they will treat the waste according to the treatment plan.

Images