A method and system for intelligent meeting scheduling of separate repair tunnels

Abstract

The application provides a kind of partial repair tunnel intelligent meeting scheduling method and system, it is related to meeting scheduling technical field, this method includes: the M image in target tunnel is collected;Obtain N target images that exist vehicle, and obtain N direction of travel and N position information;Obtain N vehicle feature set;Obtain N meeting distance;Meeting information matrix is constructed, and is normalized to standard information matrix, and N scheduling priority coefficient is calculated;The forward scheduling priority coefficient and the reverse scheduling priority coefficient are calculated, and the vehicle in the direction of smaller forward scheduling priority coefficient and the reverse scheduling priority coefficient is scheduled, the technical problem that the meeting effect is not good in the prior art, influence vehicle operation efficiency is solved, the priority analysis of vehicle meeting scheduling is realized, and then the technical effect that the meeting effect is improved, and then the influence of meeting on vehicle operation efficiency is reduced, the operation efficiency of vehicle in tunnel is improved.

Description

Technical Field

[0001] This invention relates to the field of train passing and scheduling technology, specifically to an intelligent train passing and scheduling method and system for tunnel maintenance. Background Technology

[0002] During the construction of the tunnel, vehicles need to operate in the tunnel, such as transporting ore. The lanes inside the tunnel are narrow, making it difficult for two vehicles to pass side by side. In order to maximize the use of the lanes and improve the efficiency of vehicle traffic, a meeting point is specially set up at a fixed location. When an oncoming vehicle notices an oncoming vehicle, it can wait temporarily at the meeting point, thus solving the problem of multiple vehicles passing through the tunnel in both directions.

[0003] Currently, existing technologies mostly rely on the distance between vehicles and the meeting point for scheduling when vehicles meet, without sufficient analysis of the vehicles' own operating and driving characteristics. This leads to technical problems such as poor scheduling effectiveness and reduced vehicle operating efficiency. Summary of the Invention

[0004] This invention provides an intelligent vehicle meeting scheduling method and system for tunnel maintenance, which solves the technical problem of poor vehicle meeting scheduling effect and reduced vehicle operation efficiency in the prior art.

[0005] According to a first aspect of the present invention, a method for intelligent vehicle passing scheduling in a tunnel under separate maintenance is provided, comprising: initiating a vehicle passing scheduling command; acquiring M images of M locations within a target tunnel, where M is an integer greater than 1; preprocessing and performing image feature analysis on the M images to obtain N target images containing vehicles, and analyzing and obtaining N directions of travel and N position information of the vehicles within the N target images, where N is an integer greater than 1 and less than or equal to M; determining whether the N directions of travel are consistent; if so, ending the vehicle passing scheduling command; otherwise, performing feature analysis on the N target images to obtain N identification information of the vehicles within the N target images, and inputting this information into a vehicle passing scheduling database for retrieval to obtain N vehicles. A feature set is used for each vehicle, including speed level, task level, agility level, and parking difficulty level. Based on the N location information and the coordinate set of multiple meeting points within the target tunnel, N meeting distances between vehicles in the N target images and their neighboring meeting points are calculated. Using the N vehicle feature sets and N meeting distances, a meeting information matrix is ​​constructed and normalized to a standard information matrix, and N scheduling priority coefficients are calculated. Based on the N travel directions, cluster analysis is performed on the vehicles in the N target images, and forward scheduling priority coefficients and reverse scheduling priority coefficients are calculated. Vehicles in the direction with the smaller forward scheduling priority coefficient and reverse scheduling priority coefficient are scheduled.

[0006] According to a second aspect of the present invention, a smart vehicle passing scheduling system for a tunnel under separate maintenance is provided, comprising: an image acquisition module, wherein the image acquisition module is used to initiate a vehicle passing scheduling command and acquire M images of M locations within a target tunnel, where M is an integer greater than 1; an image feature analysis module, wherein the image feature analysis module is used to preprocess and perform image feature analysis on the M images to obtain N target images containing vehicles, and analyze and obtain N directions of travel and N position information of vehicles within the N target images, where N is an integer greater than 1 and less than or equal to M; and a vehicle feature analysis module, wherein the vehicle feature analysis module is used to determine whether the N directions of travel are consistent; if so, the vehicle passing scheduling command is terminated; if not, feature analysis is performed on the N target images to obtain N identification information of vehicles within the N target images, and this information is input into a vehicle passing scheduling database for retrieval to obtain N vehicle features. The system comprises: a vehicle feature set, where each vehicle feature set includes speed level, task level, agility level, and parking difficulty level; a meeting distance calculation module, which calculates N meeting distances between vehicles and neighboring meeting points in the N target images based on the N location information and the coordinate set of multiple meeting points within the target tunnel; a scheduling priority coefficient calculation module, which constructs a meeting information matrix using the N vehicle feature sets and N meeting distances, normalizes it to a standard information matrix, and calculates N scheduling priority coefficients; and a vehicle scheduling module, which performs cluster analysis on vehicles in the N target images based on the N travel directions, calculates forward scheduling priority coefficients and reverse scheduling priority coefficients, and schedules vehicles in the direction with the smaller forward scheduling priority coefficient and reverse scheduling priority coefficient.

[0007] The beneficial effects that can be achieved by adopting one or more technical solutions according to the present invention are as follows:

[0008] By acquiring N target images of vehicles within the target tunnel and analyzing them to obtain N directions of travel and N position information of the vehicles in the N target images, the system determines whether the N directions of travel are consistent. If they are, the vehicle passing dispatch instruction is terminated; otherwise, feature analysis is performed on the N target images to obtain N identification information of the vehicles in the N target images. This information is then input into the vehicle passing dispatch database for retrieval, resulting in N vehicle feature sets. Each vehicle feature set includes speed level, task level, agility level, and parking difficulty level. Based on the N position information and the coordinate set of multiple passing points within the target tunnel, the system calculates... The method obtains N meeting distances between vehicles and nearby meeting points in N target images. Using N vehicle feature sets and N meeting distances, a meeting information matrix is ​​constructed and normalized to a standard information matrix. N scheduling priority coefficients are calculated. Based on N travel directions, cluster analysis is performed on the vehicles in the N target images, and forward scheduling priority coefficients and reverse scheduling priority coefficients are calculated. Vehicles in the direction with the smaller forward scheduling priority coefficient and reverse scheduling priority coefficient are scheduled to reduce the impact of meeting on vehicle operation efficiency, improve vehicle scheduling effect, and thus improve the technical effect of vehicle operation efficiency in tunnels.

[0009] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0011] Figure 1 A flowchart illustrating an intelligent train meeting scheduling method for tunnel maintenance provided in an embodiment of the present invention;

[0012] Figure 2 This is a flowchart illustrating the process of obtaining N vehicle feature sets in a separate maintenance tunnel intelligent vehicle meeting scheduling method according to the present invention.

[0013] Figure 3 This is a flowchart illustrating the process of obtaining N passing distances in a separate maintenance tunnel intelligent passing scheduling method according to the present invention.

[0014] Figure 4 This is a schematic diagram of a separate maintenance tunnel intelligent vehicle meeting scheduling system provided in an embodiment of the present invention.

[0015] Figure labeling: Image acquisition module 11, Image feature analysis module 12, Vehicle feature analysis module 13, Meeting distance calculation module 14, Dispatch priority coefficient calculation module 15, Vehicle dispatching module 16. Detailed Implementation

[0016] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of the present invention, including various details to aid understanding. These details should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the invention. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0017] Example 1

[0018] Figure 1 This invention provides a method for intelligent train dispatching in a tunnel under separate maintenance, comprising:

[0019] Step S100: Initiate the meeting dispatch command and acquire M images from M locations within the target tunnel, where M is an integer greater than 1;

[0020] Specifically, the intelligent vehicle passing scheduling method for tunnel maintenance provided in this embodiment of the invention is mainly used for vehicle passing scheduling when working in a single lane in a tunnel. A single lane is a lane that allows vehicles to travel in both directions, but only one vehicle can pass in each direction. When vehicles are running in two directions, it is necessary to schedule vehicles in one direction to ensure vehicle safety.

[0021] A meeting-and-traffic scheduling command is a control command used to initiate meeting-and-traffic scheduling. It can be executed periodically, such as every hour, or when meeting-and-traffic scheduling is needed, such as in real time when traffic congestion occurs. The target tunnel refers to any tunnel to be subject to meeting-and-traffic scheduling. Image acquisition devices, such as smart cameras, are pre-installed at M locations in the target tunnel. Connecting to these image acquisition devices, the meeting-and-traffic scheduling command controls the image acquisition devices installed in the target tunnel to acquire images at the M locations within the target tunnel, obtaining M images, where M is an integer greater than 1.

[0022] Step S200: Preprocess and analyze the M images to obtain N target images containing vehicles, and analyze and obtain N directions of travel and N position information of vehicles in the N target images, where N is an integer greater than 1 and less than or equal to M;

[0023] In this embodiment of the invention, step S200 further includes:

[0024] Step S210: Denoise and convert the M images to grayscale to obtain M grayscale images;

[0025] Step S220: Calculate the average gray value in the M grayscale images, and determine whether it is greater than the grayscale threshold. If it is, determine that a vehicle exists; if not, determine that no vehicle exists, and obtain the N target images of the vehicle.

[0026] The embodiments of the present invention also include:

[0027] Step S230: Crawling, analyzing and marking the monitoring images within the target tunnel over a historical period to obtain a set of sample grayscale images, a set of sample travel directions, and a set of sample distances. The sample distance set includes the sample distances between the vehicle and the image acquisition devices at M locations.

[0028] Step S240: Based on the set of sample grayscale images, the set of sample travel directions, and the set of sample distances as construction data, a convolutional neural network is used to construct a vehicle image analysis channel;

[0029] Step S250: Input the N target images after denoising and grayscale processing into the vehicle image analysis channel to obtain the N directions of travel and N distances;

[0030] Step S260: Combine the N locations within the M locations with the N distances to calculate and obtain the N location information.

[0031] Specifically, the M images are denoised and converted to grayscale to obtain M grayscale images. Image denoising methods include median filtering, mean filtering, Gaussian filtering, etc., and the appropriate image denoising method can be selected according to the actual situation. Image denoising is implemented using software platforms such as MATLAB and Python. Then, the M denoised images are converted to grayscale. Grayscale conversion is the process of converting a multi-channel color image into a single-channel grayscale image. A color image contains changes in three color channels: red (R), green (G), and blue (B). If R = G = B, then the color represents a grayscale color. This process is called image grayscale conversion. There are three common grayscale conversion methods: First, take the value of the largest value among the three components R, G, and B (0 is considered the smallest, and 255 is considered the largest); second, take the average value among the three components R, G, and B; third, obtain a weighted average based on the sensitivity of the human eye to the three colors R, G, and B according to a certain weight. In practical applications, one can choose one of the grayscale conversion methods and use software platforms such as MATLAB and Python to perform image grayscale processing, thereby obtaining M grayscale images.

[0032] The average grayscale value of each of the M grayscale images is further calculated to obtain M average grayscale values. Then, when no vehicles are passing through the target tunnel, a tunnel image without vehicles is acquired. This tunnel image undergoes image denoising and grayscale processing, and the average grayscale value of the tunnel image is calculated as a grayscale threshold. Then, it is determined whether each of the M average grayscale values ​​is greater than the grayscale threshold. If yes, it is determined that a vehicle exists in the grayscale image; otherwise, it is determined that no vehicle exists. This yields the N target images containing vehicles. Simply put, the light inside the tunnel is relatively dim, so vehicles will turn on their headlights when driving in the tunnel. Therefore, when there are no vehicles in the M grayscale images, the image grayscale is not affected by vehicle headlights, and its average grayscale value will be less than or equal to the grayscale threshold. If the average grayscale value is greater than the grayscale threshold, it indicates that a vehicle exists in the corresponding grayscale image. Based on this, the grayscale images with an average grayscale value greater than the grayscale threshold from the M grayscale images are extracted as the N target images, where N is an integer greater than 1 and less than or equal to M. This enables the determination of vehicle presence and the extraction of target images, providing a foundation for subsequent vehicle meeting and dispatching.

[0033] Furthermore, monitoring images from historical periods (e.g., the past month) within the target tunnel are crawled, analyzed, and labeled. Image crawling refers to automatically capturing monitoring images and downloading them to a local computer based on a set historical time. The monitoring images include historical vehicle travel images at M locations within the target tunnel. The obtained monitoring images are then analyzed. First, the monitoring images are denoised and converted to grayscale to obtain a set of sample grayscale images. Then, the travel direction of the vehicles in the monitoring images is labeled. Preferably, the direction of the vehicle lights can be used as the vehicle's travel direction to obtain a set of sample travel directions. Further, the distances between the vehicles in the monitoring images and the image acquisition devices at the M locations are obtained. Specifically, this can be based on M... The user manual for the image acquisition device at each location obtains the mapping relationship between the scaling factor of the acquisition target and the acquisition distance. Then, based on the vehicle model in the monitoring image, the actual size of the vehicle is determined, and the size of the vehicle in the image is obtained. Based on the actual size and the size of the vehicle in the image, the scaling factor of the monitoring image can be obtained. Then, based on the mapping relationship between the scaling factor and the acquisition distance, the acquisition distance corresponding to the scaling factor of the monitoring image can be obtained and combined to obtain a sample distance set. Thus, the sample distance set includes the sample distances between the vehicle and the image acquisition devices at M locations. Among them, the data in the sample grayscale image set, the sample travel direction set, and the sample distance set have a one-to-one correspondence.

[0034] Using the aforementioned set of sample grayscale images, sample travel directions, and sample distances as construction data, a vehicle image analysis channel is constructed using a convolutional neural network. The input to the vehicle image analysis channel is a grayscale image, and the output is the travel direction and distance. Simply put, each sample grayscale image in the sample grayscale image set is input into the vehicle image analysis channel. The output of the vehicle image analysis channel is supervised and adjusted using the corresponding sample travel directions and sample distances from the sample travel direction and sample distance sets, so that the output of the vehicle image analysis channel is consistent with the sample travel direction and sample distance. This completes the training of the vehicle image analysis channel. Then, the output accuracy of the vehicle image analysis channel is tested to obtain a vehicle image analysis channel that meets the requirements.

[0035] The N target images, after denoising and grayscale processing, are sequentially input into the vehicle image analysis channel. This outputs the N directions of travel and N distances sequentially. The N positions are subsets of the M positions. Combining the N positions within the M positions with the N distances, the N positional information is calculated. Simply put, the N positions and N distances are known conditions. The N positions refer to the coordinates of the image acquisition devices within the N positions, and the N distances are the distances between the vehicle in the N target images and the image acquisition devices within the N positions. The N positional information can be obtained through simple mathematical calculations (knowing the distance between two points and the position of one point, calculating the position of the other). This N positional information represents the vehicle's location in the N target images. This enables the analysis of vehicle positions, providing support for subsequent dispatching.

[0036] Step S300: Determine whether the N directions of travel are consistent. If yes, end the meeting dispatch instruction. If no, perform feature analysis based on the N target images to obtain N identification information of vehicles in the N target images and input them into the meeting dispatch database for retrieval to obtain N vehicle feature sets. Each vehicle feature set includes speed level, task level, flexibility level, and parking difficulty level.

[0037] Among them, such as Figure 2 As shown, step S300 of this embodiment of the invention further includes:

[0038] Step S310: Crawling the monitoring images within the target tunnel over a historical period, identifying and marking the vehicles within the monitoring images, and obtaining a set of sample grayscale images and a set of sample identification information;

[0039] Step S320: Based on the set of sample grayscale images and the set of sample identification information, a vehicle recognition channel is constructed using a convolutional neural network;

[0040] Step S330: Input the N target images after denoising and grayscale processing into the vehicle recognition channel to obtain the N identification information;

[0041] Step S340: Based on the operation information of all vehicles in the target tunnel, and combined with the sample identification information set, obtain the moving speed, task importance information, operational flexibility information and parking difficulty information of all vehicles, and allocate and obtain sample speed level set, sample task level set, sample flexibility level set and sample parking difficulty level set.

[0042] Step S350: Based on the sample identification information set, sample speed level set, sample task level set, sample flexibility level set, and sample parking difficulty level set, construct index elements and data elements to obtain the meeting scheduling database;

[0043] Step S360: Input the N identification information into the vehicle dispatch database for retrieval to obtain the N vehicle feature sets.

[0044] Specifically, it is determined whether the N directions of travel are consistent. If they are, it means that the vehicles in the target tunnel are traveling in the same direction, and no passing scheduling is needed, so the passing scheduling instruction ends. If the N directions of travel are inconsistent, feature analysis is performed on the N target images to obtain N identification information of the vehicles in the N target images, and this information is entered into the passing scheduling database for retrieval to obtain N vehicle feature sets. Each vehicle feature set includes speed level, task level, agility level, and parking difficulty level. The specific process is as follows:

[0045] The system crawls monitoring images from a historical period within the target tunnel. The historical period can be set arbitrarily, for example, the past month, downloading monitoring images from the past month. Vehicles within these images are then identified and marked. In simpler terms, the monitoring images are converted to grayscale to create a sample grayscale image set. Vehicles within the monitoring images are then identified by features such as license plates, vehicle size, and vehicle appearance. Vehicle feature recognition is a common technique used by those skilled in the art, and therefore will not be elaborated upon here. This yields a sample identification information set. Based on the sample grayscale image set and the sample identification information set... A convolutional neural network is used to construct a vehicle recognition channel. The data in the sample grayscale image set and the sample identification information set have a one-to-one correspondence. The input to the vehicle recognition channel is a grayscale image, and the output is identification information. Each sample grayscale image in the sample grayscale image set is sequentially input into the vehicle recognition channel. The output of the vehicle recognition channel is supervised and adjusted using the corresponding sample identification information in the sample identification information set, ensuring that the output of the vehicle recognition channel is consistent with the sample identification information. After training with all data in both the sample grayscale image set and the sample identification information set, the output accuracy of the trained vehicle recognition channel is tested to obtain a vehicle recognition channel with the required accuracy. The N target images after denoising and grayscale processing are sequentially input into the vehicle recognition channel, and the N identification information are output. Further, based on the operational information of all vehicles within the target tunnel, including the transportation of ore, coal, and other materials, the specific information can be obtained according to the actual situation.By combining the operational information of all vehicles with the sample identification information set, the moving speed, task importance, operational flexibility, and parking difficulty information of all vehicles are obtained. Moving speed refers to the vehicle's speed under normal operational conditions. Speed ​​levels are analyzed according to the magnitude of the moving speed. Simply put, the moving speed of all vehicles can be divided into multiple speed ranges at equal intervals, and a speed level is assigned to each speed range. The larger the speed range, the higher the speed level. This results in a sample speed level set for all vehicles. Task importance information refers to the degree of importance of the tasks performed by all vehicles, which is determined by those skilled in the art based on the actual situation. For example, transporting ore is of higher importance than transporting ordinary materials. A sample task level is assigned to each vehicle according to the task importance; the higher the task importance, the higher the corresponding sample task level. The higher the level, the more sample task level sets are constructed. The operational flexibility information can be determined based on the vehicle model in the sample identification information set. Different vehicles have different chassis, engines, transmissions, etc., so the operational flexibility naturally varies. The operational flexibility of all vehicles can be obtained based on existing technology, and a sample flexibility level is assigned to each vehicle according to the level of operational flexibility. The higher the operational flexibility, the higher the corresponding sample flexibility level, thus obtaining a sample flexibility level set. Parking difficulty is related to the vehicle's moving speed and size. The faster the vehicle moves and the larger the vehicle, the higher the corresponding parking difficulty. Based on this, the parking difficulty information of all vehicles is obtained according to the moving speed and size of all vehicles. A sample parking difficulty level is assigned to each vehicle according to the level of parking difficulty. The higher the parking difficulty, the higher the corresponding sample parking difficulty level, thus obtaining a sample parking difficulty level set.

[0046] Based on the sample identification information set, sample speed level set, sample task level set, sample flexibility level set, and sample parking difficulty level set, index elements and data elements are constructed. The index elements are the sample identification information set, and the data elements are the sample speed level set, sample task level set, sample flexibility level set, and sample parking difficulty level set. The meeting dispatch database is composed of the index elements and data elements. The N identification information is input into the meeting dispatch database for retrieval. First, the N identification information is searched in the index elements, i.e., the sample identification information set, to obtain sample identification information that is the same as the N identification information. Then, the corresponding data is retrieved in the data elements, i.e., the sample speed level set, sample task level set, sample flexibility level set, and sample parking difficulty level set, to obtain N sample speed levels, N sample task levels, N sample flexibility levels, and N sample parking difficulty levels corresponding to the N identification information. One sample speed level, one sample task level, one sample flexibility level, and one sample parking difficulty level are combined to obtain a vehicle feature set, thus obtaining N vehicle feature sets. This enables the analysis of vehicle characteristics, providing technical support for subsequent vehicle meeting and dispatching, and reducing the impact of vehicle meeting and dispatching on vehicle operation efficiency.

[0047] Step S400: Based on the N location information and the coordinate set of multiple meeting points in the target tunnel, calculate the N meeting distances between the vehicle and the adjacent meeting points in the N target images;

[0048] Among them, such as Figure 3 As shown, step S400 of this embodiment of the invention further includes:

[0049] Step S410: Obtain the position coordinates of the multiple meeting points within the target tunnel, and combine them to obtain the set of meeting point coordinates;

[0050] Step S420: According to the threshold for the number of vehicles meeting at the multiple meeting points, the number of vehicles meeting at each meeting point is not greater than the threshold for the number of vehicles meeting.

[0051] Step S430: Based on the constraints, according to the N location information and the set of meeting point coordinates, assign the vehicles in the N target images to the nearest meeting points to obtain N assignment results;

[0052] Step S440: Based on the N allocation results, calculate the distances between the vehicles in the N target images and the allocated meeting points to obtain the N meeting distances.

[0053] Specifically, tunnels typically have multiple passing points for traffic scheduling. The locations of these passing points are fixed and known. Based on this, the coordinates of these multiple passing points within the target tunnel can be directly read from the tunnel design drawings, forming the passing point coordinate set. A passing number threshold is set, ensuring that the number of vehicles passing at each passing point does not exceed this threshold. This threshold is determined by those skilled in the art based on actual conditions. The passing number threshold represents the allowed number of vehicles passing at each passing point, i.e., the number of vehicles that can park there. It is pre-designed during tunnel construction and can be directly read.

[0054] Based on the constraints, and according to the N location information and the set of meeting point coordinates, vehicles in the N target images are assigned to nearby meeting points. Simply put, the distances between vehicles in the N target images and multiple meeting points are calculated based on the N location information and the set of meeting point coordinates. Then, the meeting point closest to each vehicle in the target image is determined as the assignment result for that vehicle. This process is repeated to obtain N assignment results. Based on the N assignment results, the distances between the vehicles in the N target images and the assigned meeting points are calculated. In simpler terms, the N assignment results are the N meeting points corresponding to each vehicle in the target image. There may be shared meeting points among the N meeting points. Since the positions of the meeting points and vehicles are known, the distances between the vehicles in the N target images and the assigned meeting points can be obtained through simple distance calculations. These distances are then used as the N meeting distances, thus enabling vehicle meeting and facilitating subsequent vehicle meeting scheduling based on these distances, thereby improving scheduling efficiency.

[0055] Step S500: Using the N vehicle feature sets and N meeting distances, construct a meeting information matrix, normalize it into a standard information matrix, and calculate N scheduling priority coefficients;

[0056] In this embodiment of the invention, step S500 further includes:

[0057] Step S510: Maximize the N flexibility level data within the N vehicle feature sets;

[0058] Step S520: Based on the processed N vehicle feature sets and N meeting distances, construct the meeting information matrix as follows:

[0059]

[0060] in, Let the speed level of the vehicle be the first image among N target images. Let N be the speed level of the vehicle in the Nth target image out of N target images. Let be the passing distance between the vehicle in the first image among N target images and its nearest passing point. The passing distance between a vehicle in the Nth target image and its nearest passing point is given by the Nth target image.

[0061] Step S530: Normalize the elements in the meeting information matrix to obtain a standard information matrix, as shown below:

[0062]

[0063]

[0064] in, for The standard value after normalization.

[0065] Specifically, the N flexibility level data within the N vehicle feature sets are maximized. Maximization is a basic data standardization method. The maximum value of the N flexibility level data is taken, and then divided by each flexibility level data. This means that vehicles with low speed, low task, high flexibility, and low parking difficulty levels need to stop at the meeting platform to wait for the vehicles to pass. In other words, the flexibility level is not aligned with the other four parameters. Maximization reverses the order of the N flexibility level data by dividing the maximum value by each flexibility level data, making the largest value the smallest value. This aligns the speed, task, flexibility, and parking difficulty levels, facilitating subsequent calculation of scheduling priority coefficients. Based on the processed N vehicle feature sets and N meeting distances, the meeting information matrix is ​​constructed as follows:

[0066]

[0067] in, Let the speed level of the vehicle be the first image among N target images. Let N be the speed level of the vehicle in the Nth target image out of N target images. Let be the passing distance between the vehicle in the first image among N target images and its nearest passing point. The meeting distance between a vehicle in the Nth target image and its nearest meeting point is defined as follows: In simple terms, each row of the meeting information matrix represents the speed level, task level, agility level, parking difficulty level, and meeting distance of a vehicle in the target image. Each column represents the speed level, task level, agility level, parking difficulty level, and meeting distance of all vehicles. The meeting information matrix has a fixed number of 5 columns and N rows.

[0068] The elements within the meeting information matrix are normalized to obtain a standard information matrix, as shown in the following formula:

[0069]

[0070]

[0071] in, for The normalized standard value can be used to normalize other elements using the same method.

[0072] In simple terms, the method involves normalizing the elements in the vehicle passing information matrix using square normalization. This is done by calculating the sum of squares of all elements in the matrix, then calculating the square root of the sum, and finally using the ratio of each element to the square root of the sum as the standard value after normalization. It should be noted that there are various normalization methods available, and those skilled in the art can choose other normalization methods (such as centering or mean normalization) based on the specific circumstances. There are no restrictions on this approach.

[0073] Based on the standard information matrix, the N scheduling priority coefficients are calculated as follows:

[0074]

[0075]

[0076]

[0077] Among them, K i w is the scheduling priority coefficient for vehicles within the i-th target image. j To assign weights to the scheduling priority coefficient based on the j-th parameter among speed level, task level, flexibility level, parking difficulty level, and meeting distance, preferably, the weight coefficients of speed level, task level, flexibility level, parking difficulty level, and meeting distance can be analyzed separately using the coefficient of variation method, thereby obtaining the weights corresponding to speed level, task level, flexibility level, parking difficulty level, and meeting distance respectively, max j This refers to the maximum value among the standard values ​​in the j-th column of the standard information matrix, which is the maximum value of any parameter (speed level, task level, agility level, parking difficulty level, and meeting distance) corresponding to the vehicles in the N target images. For example, the maximum value among the speed levels corresponding to the vehicles in the N target images, min j This refers to the minimum value among the standard values ​​in the j-th column of the standard information matrix, which is the minimum value of any parameter (speed level, task level, agility level, parking difficulty level, and meeting distance) corresponding to the vehicle in each of the N target images. This represents the standard value in the i-th row and j-th column of the standard information matrix. This allows for the calculation of the scheduling priority coefficients for vehicles within N target images, providing support for vehicle meeting scheduling.

[0078] Step S600: Based on the N travel directions, perform cluster analysis on the vehicles in the N target images, calculate the forward scheduling priority coefficient and the reverse scheduling priority coefficient, and schedule the vehicles in the direction with the smaller forward scheduling priority coefficient and the reverse scheduling priority coefficient.

[0079] Specifically, based on the N travel directions, cluster analysis is performed on the vehicles in the N target images. Simply put, the N travel directions include two categories: those entering the tunnel and those exiting the tunnel. Based on this, the vehicles in the N target images are divided into two categories: vehicles entering the tunnel and vehicles exiting the tunnel. Then, the scheduling priority coefficients corresponding to vehicles entering the tunnel are summed to obtain a forward scheduling priority coefficient; the scheduling priority coefficients corresponding to vehicles exiting the tunnel are summed to obtain a reverse scheduling priority coefficient. The forward and reverse scheduling priority coefficients represent the scheduling priority coefficients of vehicles in two opposite directions. Comparing the forward and reverse scheduling priority coefficients, vehicles in the direction with the smaller forward and reverse scheduling priority coefficients are scheduled. In simpler terms, a meeting point scheduling command is sent to the vehicle in the direction with the smaller forward and reverse scheduling priority coefficients, controlling the vehicle in the smaller direction to travel to the meeting point in the N allocation results and stop to wait, thus completing the meeting. This enables the scheduling of passing vehicles within the target tunnel, thereby reducing the impact of passing vehicle scheduling on vehicle operations and improving the efficiency of vehicle operations within the tunnel.

[0080] Based on the above analysis, the beneficial effects that can be achieved by one or more technical solutions adopted according to the present invention are as follows:

[0081] By acquiring N target images of vehicles within the target tunnel and analyzing them to obtain N directions of travel and N position information of the vehicles in the N target images, the system determines whether the N directions of travel are consistent. If they are, the vehicle passing dispatch instruction is terminated; otherwise, feature analysis is performed on the N target images to obtain N identification information of the vehicles in the N target images. This information is then input into the vehicle passing dispatch database for retrieval, resulting in N vehicle feature sets. Each vehicle feature set includes speed level, task level, agility level, and parking difficulty level. Based on the N position information and the coordinate set of multiple passing points within the target tunnel, the system calculates... The method obtains N meeting distances between vehicles and nearby meeting points in N target images. Using N vehicle feature sets and N meeting distances, a meeting information matrix is ​​constructed and normalized to a standard information matrix. N scheduling priority coefficients are calculated. Based on N travel directions, cluster analysis is performed on the vehicles in the N target images, and forward scheduling priority coefficients and reverse scheduling priority coefficients are calculated. Vehicles in the direction with the smaller forward scheduling priority coefficient and reverse scheduling priority coefficient are scheduled to reduce the impact of meeting on vehicle operation efficiency, improve vehicle scheduling effect, and thus improve the technical effect of vehicle operation efficiency in tunnels.

[0082] Example 2

[0083] Based on the same inventive concept as the intelligent train meeting scheduling method for split-maintenance tunnels described in the foregoing embodiments, such as Figure 4 As shown, the present invention also provides an intelligent vehicle meeting scheduling system for tunnel maintenance, the system comprising:

[0084] Image acquisition module 11, which is used to initiate a meeting dispatch command and acquire M images at M locations within the target tunnel, where M is an integer greater than 1;

[0085] Image feature analysis module 12 is used to preprocess and analyze the M images to obtain N target images containing vehicles, and to analyze and obtain N directions of travel and N position information of vehicles in the N target images, where N is an integer greater than 1 and less than or equal to M.

[0086] The vehicle feature analysis module 13 is used to determine whether the N directions of travel are consistent. If they are, the meeting dispatch instruction ends. If not, feature analysis is performed on the N target images to obtain N identification information of the vehicles in the N target images. The information is then entered into the meeting dispatch database for retrieval to obtain N vehicle feature sets. Each vehicle feature set includes speed level, task level, flexibility level, and parking difficulty level.

[0087] The vehicle meeting distance calculation module 14 is used to calculate the N vehicle meeting distances between the vehicle and the adjacent vehicle meeting point in the N target images based on the N location information and the coordinate set of multiple vehicle meeting points in the target tunnel.

[0088] The scheduling priority coefficient calculation module 15 is used to construct a meeting information matrix using the N vehicle feature sets and N meeting distances, and normalize it into a standard information matrix to calculate N scheduling priority coefficients.

[0089] The vehicle scheduling module 16 is used to perform cluster analysis on vehicles in the N target images according to the N directions of travel, calculate the forward scheduling priority coefficient and the reverse scheduling priority coefficient, and schedule vehicles in the direction with the smaller forward scheduling priority coefficient and the reverse scheduling priority coefficient.

[0090] Furthermore, the image feature analysis module 12 is also used for:

[0091] The M images are denoised and converted to grayscale to obtain M grayscale images;

[0092] Calculate the average grayscale value within the M grayscale images, and determine whether it is greater than the grayscale threshold. If it is, determine that a vehicle exists; otherwise, determine that no vehicle exists, and obtain the N target images containing the vehicle.

[0093] Furthermore, the image feature analysis module 12 is also used for:

[0094] The monitoring images within the target tunnel over a historical period are crawled, analyzed, and labeled to obtain a set of sample grayscale images, a set of sample travel directions, and a set of sample distances. The sample distance set includes the sample distances between the vehicle and the image acquisition devices at M locations.

[0095] Based on the set of grayscale images, the set of travel directions, and the set of distances, a vehicle image analysis channel is constructed using a convolutional neural network.

[0096] The N target images after denoising and grayscale processing are input into the vehicle image analysis channel to obtain the N travel directions and N distances;

[0097] By combining the N locations within the M locations and the N distances, the N location information is calculated.

[0098] Furthermore, the vehicle feature analysis module 13 is also used for:

[0099] The monitoring images within the target tunnel over a historical period are crawled, and vehicles within the monitoring images are identified and marked to obtain a set of sample grayscale images and a set of sample identification information.

[0100] Based on the set of sample grayscale images and the set of sample identification information, a vehicle recognition channel is constructed using a convolutional neural network.

[0101] The N target images after denoising and grayscale processing are input into the vehicle recognition channel to obtain the N identification information;

[0102] Based on the operational information of all vehicles in the target tunnel, and combined with the sample identification information set, the moving speed, operational task importance information, operational flexibility information, and parking difficulty information of all vehicles are obtained, and sample speed level set, sample task level set, sample flexibility level set, and sample parking difficulty level set are allocated and obtained.

[0103] Based on the sample identification information set, sample speed level set, sample task level set, sample flexibility level set, and sample parking difficulty level set, index elements and data elements are constructed to obtain the meeting scheduling database.

[0104] The N identification information is input into the vehicle meeting scheduling database for retrieval to obtain the N vehicle feature sets.

[0105] Furthermore, the meeting distance calculation module 14 is also used for:

[0106] Obtain the position coordinates of the multiple meeting points within the target tunnel, and then combine them to obtain the set of meeting point coordinates.

[0107] Based on the threshold for the number of vehicles meeting at the multiple meeting points, the constraint condition is that the number of vehicles meeting at each meeting point is not greater than the threshold for the number of vehicles meeting.

[0108] Based on the constraints, and according to the N location information and the set of meeting point coordinates, the vehicles in the N target images are assigned to nearby meeting points to obtain N assignment results.

[0109] Based on the N allocation results, the distances between the vehicles in the N target images and the allocated meeting points are calculated to obtain the N meeting distances.

[0110] Furthermore, the scheduling priority coefficient calculation module 15 is also used for:

[0111] Maximize the N flexibility level data within the N vehicle feature sets;

[0112] Based on the processed N vehicle feature sets and N meeting distances, the meeting information matrix is ​​constructed as follows:

[0113]

[0114] in, Let the speed level of the vehicle be the first image among N target images. Let N be the speed level of the vehicle in the Nth target image out of N target images. Let be the passing distance between the vehicle in the first image among N target images and its nearest passing point. The passing distance between a vehicle in the Nth target image and its nearest passing point is given by the Nth target image.

[0115] The elements within the meeting information matrix are normalized to obtain a standard information matrix, as shown in the following formula:

[0116]

[0117]

[0118] in, for The standard value after normalization.

[0119] Furthermore, the scheduling priority coefficient calculation module 15 is also used for:

[0120] Based on the standard information matrix, the N scheduling priority coefficients are calculated as follows:

[0121]

[0122]

[0123]

[0124] Among them, K i w is the scheduling priority coefficient for vehicles within the i-th target image. j The weights assigned to the scheduling priority coefficient based on the j-th parameter among speed level, task level, flexibility level, parking difficulty level, and meeting distance are max. j min represents the maximum value among the standard values ​​in the j-th column of the standard information matrix. j The minimum value among the standard values ​​in the j-th column of the standard information matrix. is the standard value in the i-th row and j-th column of the standard information matrix.

[0125] The specific example of the intelligent train dispatching method for split-maintenance tunnels in the aforementioned Embodiment 1 is also applicable to the intelligent train dispatching system for split-maintenance tunnels in this embodiment. Through the detailed description of the intelligent train dispatching method for split-maintenance tunnels described above, those skilled in the art can clearly understand the intelligent train dispatching system for split-maintenance tunnels in this embodiment. Therefore, for the sake of brevity, it will not be described in detail here.

[0126] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this invention can be achieved, and this is not limited herein.

[0127] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for intelligent train meeting scheduling in a tunnel under separate maintenance, characterized in that, The method includes: Initiate the meeting dispatch command and collect M images from M locations within the target tunnel, where M is an integer greater than 1; The M images are preprocessed and their features are analyzed to obtain N target images containing vehicles. The N travel directions and N position information of the vehicles in the N target images are then analyzed and obtained, where N is an integer greater than 1 and less than or equal to M. Determine whether the N directions of travel are consistent. If they are, end the meeting dispatch instruction. If not, perform feature analysis based on the N target images to obtain N identification information of the vehicles in the N target images. Input the information into the meeting dispatch database for retrieval to obtain N vehicle feature sets. Each vehicle feature set includes speed level, task level, flexibility level, and parking difficulty level. Based on the N location information and the coordinate set of multiple meeting points within the target tunnel, the N meeting distances between the vehicle and the adjacent meeting points in the N target images are calculated. Using the N vehicle feature sets and N meeting distances, a meeting information matrix is ​​constructed and normalized to a standard information matrix, and N scheduling priority coefficients are calculated. Based on the N travel directions, cluster analysis is performed on the vehicles in the N target images, and forward scheduling priority coefficients and reverse scheduling priority coefficients are calculated. Vehicles in the direction with the smaller forward scheduling priority coefficient and reverse scheduling priority coefficient are scheduled.

2. The method according to claim 1, characterized in that, Preprocessing and image feature analysis are performed on the M images to obtain N target images containing vehicles, including: The M images are denoised and converted to grayscale to obtain M grayscale images; Calculate the average grayscale value within the M grayscale images, and determine whether it is greater than the grayscale threshold. If it is, determine that a vehicle exists; otherwise, determine that no vehicle exists, and obtain the N target images containing the vehicle.

3. The method according to claim 2, characterized in that, The analysis obtains N directions of travel and N position information of vehicles within the N target images, including: The monitoring images within the target tunnel over a historical period are crawled, analyzed, and labeled to obtain a set of sample grayscale images, a set of sample travel directions, and a set of sample distances. The sample distance set includes the sample distances between the vehicle and the image acquisition devices at M locations. Based on the set of grayscale images, the set of travel directions, and the set of distances, a vehicle image analysis channel is constructed using a convolutional neural network. The N target images after denoising and grayscale processing are input into the vehicle image analysis channel to obtain the N travel directions and N distances; By combining the N locations within the M locations and the N distances, the N location information is calculated.

4. The method according to claim 1, characterized in that, Based on the N target images, feature analysis is performed to obtain N identification information of vehicles within the N target images, and this information is then input into the traffic dispatch database for retrieval to obtain N vehicle feature sets, including: The monitoring images within the target tunnel over a historical period are crawled, and vehicles within the monitoring images are identified and marked to obtain a set of sample grayscale images and a set of sample identification information. Based on the set of sample grayscale images and the set of sample identification information, a vehicle recognition channel is constructed using a convolutional neural network. The N target images after denoising and grayscale processing are input into the vehicle recognition channel to obtain the N identification information; Based on the operational information of all vehicles in the target tunnel, and combined with the sample identification information set, the moving speed, operational task importance information, operational flexibility information, and parking difficulty information of all vehicles are obtained, and sample speed level set, sample task level set, sample flexibility level set, and sample parking difficulty level set are allocated and obtained. Based on the sample identification information set, sample speed level set, sample task level set, sample flexibility level set, and sample parking difficulty level set, index elements and data elements are constructed to obtain the meeting scheduling database. The N identification information is input into the vehicle meeting scheduling database for retrieval to obtain the N vehicle feature sets.

5. The method according to claim 1, characterized in that, Based on the N location information and combined with the coordinate set of multiple meeting points within the target tunnel, the N meeting distances between the vehicle and its nearest meeting point in the N target images are calculated, including: Obtain the position coordinates of the multiple meeting points within the target tunnel, and then combine them to obtain the set of meeting point coordinates. Based on the threshold for the number of vehicles meeting at the multiple meeting points, the constraint condition is that the number of vehicles meeting at each meeting point is not greater than the threshold for the number of vehicles meeting. Based on the constraints, and according to the N location information and the set of meeting point coordinates, the vehicles in the N target images are assigned to nearby meeting points to obtain N assignment results. Based on the N allocation results, the distances between the vehicles in the N target images and the allocated meeting points are calculated to obtain the N meeting distances.

6. The method according to claim 1, characterized in that, Using the N vehicle feature sets and N meeting distances, a meeting information matrix is ​​constructed and normalized to a standard information matrix, including: Maximize the N flexibility level data within the N vehicle feature sets; Based on the processed N vehicle feature sets and N meeting distances, the meeting information matrix is ​​constructed as follows: ； in, Let the speed level of the vehicle be the first image among N target images. Let N be the speed level of the vehicle in the Nth target image out of N target images. Let be the passing distance between the vehicle in the first image among N target images and its nearest passing point. The meeting distance between a vehicle in the Nth target image and its neighboring meeting points is defined in the meeting information matrix. Each row of the matrix represents the speed level, task level, agility level, parking difficulty level, and meeting distance of a vehicle in a target image. Each column represents the speed level, task level, agility level, parking difficulty level, and meeting distance of all vehicles. The elements within the meeting information matrix are normalized to obtain a standard information matrix, as shown in the following formula: ； ； in, for The normalized standard value is then used to normalize other elements using the same method.

7. The method according to claim 6, characterized in that, Calculate and obtain N scheduling priority coefficients, including: Based on the standard information matrix, the N scheduling priority coefficients are calculated as follows: ； ； ； in, Let be the scheduling priority coefficient for vehicles within the i-th target image. To assign weights to the scheduling priority coefficient based on the j-th parameter among speed level, task level, flexibility level, parking difficulty level, and meeting distance, the coefficient of variation method is used to analyze the weight coefficients of speed level, task level, flexibility level, parking difficulty level, and meeting distance respectively, thereby obtaining the weights corresponding to each of these parameters. The maximum value among the standard values ​​in the j-th column of the standard information matrix. The minimum value among the standard values ​​in the j-th column of the standard information matrix. is the standard value in the i-th row and j-th column of the standard information matrix.

8. A smart train dispatching system for tunnel maintenance, characterized in that, The system comprises the steps for performing any one of the intelligent train meeting scheduling methods for split-maintenance tunnels as described in claims 1 to 7, wherein the system includes: The image acquisition module is used to initiate a meeting dispatch command and acquire M images at M locations within the target tunnel, where M is an integer greater than 1. The image feature analysis module is used to preprocess and analyze the M images to obtain N target images containing vehicles, and to analyze and obtain N directions of travel and N position information of the vehicles in the N target images, where N is an integer greater than 1 and less than or equal to M. The vehicle feature analysis module is used to determine whether the N directions of travel are consistent. If they are, the vehicle meeting dispatch instruction is terminated. If not, feature analysis is performed on the N target images to obtain N identification information of the vehicles in the N target images. The information is then input into the vehicle meeting dispatch database for retrieval to obtain N vehicle feature sets. Each vehicle feature set includes speed level, task level, agility level, and parking difficulty level. The vehicle-to-vehicle distance calculation module is used to calculate the N vehicle-to-vehicle distances between the vehicle and the adjacent vehicle-to-vehicle points in the N target images based on the N location information and the coordinate set of multiple vehicle-to-vehicle points in the target tunnel. The scheduling priority coefficient calculation module is used to construct a meeting information matrix using the N vehicle feature sets and N meeting distances, and normalize it into a standard information matrix to calculate N scheduling priority coefficients. The vehicle scheduling module is used to perform cluster analysis on vehicles in the N target images according to the N directions of travel, calculate the forward scheduling priority coefficient and the reverse scheduling priority coefficient, and schedule vehicles in the direction with the smaller forward scheduling priority coefficient and the reverse scheduling priority coefficient.

Images