Anchor hole positioning method and system based on neural network and grey wolf optimization algorithm

Abstract

The invention relates to the technical field of machine vision and image processing, and particularly provides an anchor hole positioning method and system based on a neural network and a grey wolf optimization algorithm, and the method comprises the steps: obtaining a two-dimensional image of the inner wall of a tunnel; uniformly segmenting the two-dimensional image of the inner wall of the tunnel to obtain a plurality of square areas, and establishing a coordinate system; judging whether the preset anchor bolt hole of the tunnel contact network avoids unfavorable conditions such as cracks and water seepage points, and if so, determining the position of the current anchor bolt hole; if not, adopting a grey wolf optimization algorithm to randomly generate an artificial wolf in the square area in which the coordinate system is established; a fitness function is established, wolf group sorting is carried out, and a head wolf, a probe wolf and a fierce wolf are obtained; constructing a feedforward neural network, and obtaining the position of the head wolf and the fitness function value of the head wolf in combination with the smart behaviors of the head wolf, the probe wolf and the fierce wolf; wherein the position of the first wolf is the optimal position of the current anchor bolt hole, and the positioning of the anchor bolt hole of the tunnel contact network is completed. The invention provides a scientific solution for positioning the anchor bolt hole of the tunnel contact net.

Description

Technical Field

[0001] The present invention belongs to the technical field of machine vision and image processing, and in particular relates to an anchor hole positioning method and system based on a neural network and a gray wolf optimization algorithm. Background Art

[0002] With the rapid development of railway electrification construction, the optimization of the target position of the vacant anchor holes of the railway tunnel contact network has become a key link to ensure the stability and construction accuracy of the contact network. Tunnel contact network anchor holes refer to holes drilled on the top wall of the tunnel, which are used to fix the suspension device of the contact network system. The rationality of their layout directly affects the geometry, stability and operational safety of the contact network. However, traditional target hole location positioning methods, such as manual measurement and rule-based optimization methods, often rely on the experience of engineers to design when determining the optimal location of anchor holes. Although it can meet the basic construction needs, when facing complex tunnel structures such as irregular cross-sections, complex stress distribution, structural joints, cracks, seepage points, etc., there are many problems such as low construction efficiency, increased construction costs, poor adaptability to complex constraints, and low hole location positioning accuracy.

[0003] The existing methods for locating the anchor bolt holes of tunnel contact networks mainly adopt manual design or rule-driven optimization methods, which have the following disadvantages:

[0004] (1) High consumption of manpower and material resources: The manual design method needs to rely on the experience of engineers to optimize the design of the anchor hole position. Faced with complex tunnel top structural seams, cracks, water seepage points and irregular mechanical constraints, the design and construction efficiency is low and it is difficult to obtain the global optimal solution.

[0005] (2) Poor adaptability and easy to fall into local optimal solutions: In a complex search space, the gray wolf optimization algorithm is prone to fall into local optimal solutions when the population individuals are overly focused around the position of the alpha wolf. This phenomenon is more obvious when the search space is more complex and there are multiple local optimal solutions.

[0006] (3) Low search efficiency: The step size of the gray wolf optimization algorithm is usually fixed or lacks an effective adaptive mechanism. In the early stage of the search, the fixed step size may not be able to quickly cover the entire search space, resulting in low exploration efficiency. In the later stage of the search, the fixed step size may not be able to perform a detailed local search, resulting in a slow convergence speed.

[0007] Machine vision is widely used in environmental detection due to its advantages of high efficiency, high precision and strong adaptability. Machine vision is used to collect images of the top wall of the railway tunnel, perform image segmentation, and complete the modeling of the top environment of the tunnel. Grey Wolf Optimizer (GWO) has received widespread attention in recent years due to its strong global search capability and flexibility. Grey Wolf Optimizer is an intelligent optimization algorithm inspired by the hunting behavior of gray wolf groups. Search and hunting are completed by simulating the division of labor and cooperation among the alpha wolf, the scout wolf and the fierce wolf. However, the original Grey Wolf Optimizer is not only inefficient in dealing with target search problems, but also prone to local optimality and other problems. To this end, combined with the construction requirements of the target position positioning of the railway tunnel contact network anchor hole, the present invention proposes a method and system for positioning the tunnel contact network anchor hole based on the fusion of neural network and Grey Wolf Optimizer, so as to achieve efficient positioning of the optimal position of the optimal anchor hole.

[0008] Although the Gray Wolf Optimization Algorithm has achieved certain results in the field of target search and optimization, its performance is still significantly affected by the algorithm parameters. In particular, the parameters such as the walking step length, the rush step length, and the attack step length, which are set during the search process, will directly affect the algorithm's search ability, convergence speed, and the distribution quality of the solution. However, the traditional step length parameter settings are mostly fixed values ​​or manual adjustments based on experience, which lack adaptability and are difficult to adapt to the complex and changeable tunnel top environment.

[0009] In recent years, the development of neural network technology has provided a new solution for parameter adjustment of optimization algorithms. Neural networks have powerful nonlinear modeling capabilities and can achieve adaptive adjustment of parameters by learning the characteristic laws in data. Therefore, combining neural networks with wolf pack optimization algorithms and dynamically adjusting step size parameters through training models can significantly improve the search efficiency and accuracy of wolf pack optimization algorithms. Summary of the invention

[0010] In order to solve the problems existing in the prior art, the present invention provides an anchor hole positioning method and system based on a neural network and a gray wolf optimization algorithm. Combined with the adaptive learning ability of the neural network, the step size can be dynamically adjusted to improve the search performance for the optimal position of the anchor hole in a complex tunnel top environment.

[0011] To achieve the above object, the present invention provides the following solutions:

[0012] The anchor hole positioning method based on neural network and gray wolf optimization algorithm includes the following steps:

[0013] Scan the inner wall of the tunnel based on machine vision technology to obtain a two-dimensional image of the inner wall of the tunnel;

[0014] Evenly dividing the two-dimensional image of the inner wall of the tunnel to obtain a number of square areas, and establishing a coordinate system;

[0015] Based on the square area with the coordinate system, it is determined whether the preset anchor holes of the tunnel contact network avoid cracks and water seepage points. If so, the current anchor hole position is determined; if not, the gray wolf optimization algorithm is used to randomly generate artificial wolves in the square area with the coordinate system; wherein each artificial wolf position represents an anchor hole position;

[0016] Based on the preset objective function and the current position of the artificial wolf, a fitness function is established;

[0017] Sorting the wolf pack based on the fitness function to obtain the alpha wolf, the scout wolf and the fierce wolf;

[0018] A feedforward neural network is constructed, and the intelligent behaviors of the alpha wolf, scout wolf and fierce wolf are combined to obtain the position of the alpha wolf and the fitness function value of the alpha wolf; wherein the position of the alpha wolf is the optimal position of the current anchor hole, and the positioning of the anchor hole of the tunnel contact network is completed.

[0019] Preferably, the method for obtaining the fitness function includes:

[0020] Based on the coordinate system, the position coordinates of the artificial wolf are obtained, and based on the position coordinates, the position of the artificial wolf is obtained. The calculation formula of the position of the artificial wolf is as follows:

[0021] ,

[0022] Among them, among them, Indicates The location coordinates of the artificial wolf , and are the upper and lower bounds of the spatial search range, and the position of each artificial wolf ;

[0023] Based on the position and weight coefficient of the artificial wolf, the fitness function is obtained; wherein the weight coefficient includes a first weight coefficient for adjusting the influence ratio of the benefit in the fitness, a second weight coefficient for adjusting the proportion of the energy loss in the fitness, and a third weight coefficient for adjusting the proportion of the risk cost; the fitness function is expressed as follows:

[0024] ,

[0025] in, Represents the objective function value, measuring the current position Optimization effect; is the first weight coefficient, Indicates the current position of the artificial wolf With preset position distance; The second weight coefficient, is represented as a set of danger zones, is the indicator function, is the third weight coefficient.

[0026] Preferably, the intelligent behaviors include alpha wolf generation rules, scout wolf wandering behaviors, and alpha wolf summoning fierce wolves behaviors.

[0027] Preferably, the method for obtaining the position of the alpha wolf and the fitness function value of the alpha wolf includes:

[0028] An artificial wolf with an objective function value that meets the preset optimal requirements is selected as the leader wolf;

[0029] The direction in which the objective function value is greater than the objective function value of the current position is taken as the wandering direction; based on the wandering direction and the preset wandering step length, the scout wolf repeatedly performs the wandering behavior until the objective function value perceived by the scout wolf is greater than the objective function value of the current alpha wolf or the scout wolf reaches a preset wandering number, then the scout wolf is updated to the alpha wolf and a summoning behavior is initiated;

[0030] In the behavior of the alpha wolf calling the fierce wolf, the alpha wolf guides the fierce wolf to rush towards the alpha wolf's current position with a preset rushing step length, and obtains a feature vector based on the obstacle density, spatial scale and distance between the fierce wolf and the alpha wolf in the rushing path of the fierce wolf and the alpha wolf;

[0031] Using the feature vector as an input vector of the feedforward neural network to obtain a predicted running step length;

[0032] The fierce wolf charges based on the predicted charging step length. When the objective function value perceived by the fierce wolf is greater than the objective function value of the current alpha wolf or the fierce wolf reaches a preset charging condition, the fierce wolf is updated to the alpha wolf, and the updated position of the alpha wolf and the fitness function of the alpha wolf are obtained.

[0033] Preferably, the feedforward neural network includes an input layer, a hidden layer and an output layer;

[0034] The input layer includes three nodes, corresponding to the obstacle density, the spatial scale, and the distance between the fierce wolf and the leader wolf respectively;

[0035] The hidden layer consists of 2 layers, each containing 4 neurons;

[0036] The output layer includes 1 node, which is used to output the predicted running step length.

[0037] The present invention also provides an anchor hole positioning system based on a neural network and a gray wolf optimization algorithm, which is used to implement the method, comprising:

[0038] A two-dimensional image acquisition module is used to scan the inner wall of the tunnel based on machine vision technology to obtain a two-dimensional image of the inner wall of the tunnel;

[0039] An image segmentation module is used to evenly segment the two-dimensional image of the inner wall of the tunnel to obtain a number of square areas and establish a coordinate system;

[0040] The artificial wolf generation module is used to determine whether the preset anchor holes of the tunnel contact network avoid cracks and water seepage points based on the square area with the coordinate system. If so, the current anchor hole position is determined; if not, the gray wolf optimization algorithm is used to randomly generate artificial wolves in the square area with the coordinate system; wherein each artificial wolf position represents an anchor hole position;

[0041] A fitness function building module is used to build a fitness function based on a preset objective function and the current position of the artificial wolf;

[0042] A wolf pack sorting module is used to sort the wolf pack based on the fitness function to obtain the alpha wolf, the scout wolf and the fierce wolf;

[0043] The optimal anchor hole positioning module is used to construct a feedforward neural network, combining the intelligent behaviors of the alpha wolf, the scout wolf and the fierce wolf to obtain the position of the alpha wolf and the fitness function value of the alpha wolf; wherein the position of the alpha wolf is the optimal position of the current anchor hole, completing the positioning of the anchor hole of the tunnel contact network.

[0044] Preferably, the fitness function building module includes:

[0045] The artificial wolf position acquisition unit is used to obtain the position coordinates of the artificial wolf based on the coordinate system, and obtain the position of the artificial wolf based on the position coordinates. The calculation formula of the artificial wolf position is as follows:

[0046] ,

[0047] Among them, among them, Indicates The location coordinates of the artificial wolf , and are the upper and lower bounds of the spatial search range, and the position of each artificial wolf ;

[0048] The fitness function calculation unit is used to obtain the fitness function based on the position and weight coefficient of the artificial wolf; wherein the weight coefficient includes a first weight coefficient for adjusting the influence ratio of the benefit in the fitness, a second weight coefficient for adjusting the proportion of the energy loss in the fitness, and a third weight coefficient for adjusting the proportion of the risk cost; the fitness function is expressed as follows:

[0049] ,

[0050] in, Represents the objective function value, measuring the current position Optimization effect; is the first weight coefficient, Indicates the current position of the artificial wolf With preset position distance; The second weight coefficient, is represented as a set of danger zones, is the indicator function, is the third weight coefficient.

[0051] Preferably, the intelligent behaviors include alpha wolf generation rules, scout wolf wandering behaviors, and alpha wolf summoning fierce wolves behaviors.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] (1) Saving manpower and material resources: Based on the collected and analyzed image feature information of the tunnel top, the present invention converts the image information into an accurate mathematical model and algorithm parameters, quickly and accurately generates multiple potential anchor hole location optimization schemes, and selects the global optimal solution from them, thereby greatly improving the design and construction efficiency and positioning accuracy, and reducing the errors and risks caused by insufficient human experience.

[0054] (2) Improved adaptability: The present invention can fully adapt to the complex and changeable structural environment of the tunnel top and effectively prevent the population from converging to the local optimal solution too early, thereby ensuring that the truly global optimal anchor hole layout can be stably found in the complex search space.

[0055] (3) Improving search efficiency: With the help of the dynamic output step size of the neural network, the defects of the traditional algorithm in different search stages are effectively solved, the flexibility and accuracy of the search process are enhanced, and the search efficiency and convergence speed of the algorithm in the process of locating the position of the tunnel anchor hole are greatly improved, thus ensuring the high quality and reliability of the positioning results.

[0056] Through the above improvements, the present invention can significantly improve the global search capability, distribution uniformity and adaptability of the algorithm when solving complex optimization problems, and provide efficient optimization tools for fields such as engineering design and resource allocation. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0058] Figure 1This is a flow chart of an anchor hole positioning method based on a neural network and a gray wolf optimization algorithm according to an embodiment of the present invention;

[0059] Figure 2 A schematic diagram of randomly generating an artificial wolf according to an embodiment of the present invention. DETAILED DESCRIPTION

[0060] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0061] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0062] Embodiment 1

[0063] like Figure 1 As shown, the anchor hole positioning method based on neural network and gray wolf optimization algorithm includes the following steps:

[0064] S1: Scan the inner wall of the tunnel based on machine vision technology to obtain a two-dimensional image of the inner wall of the tunnel. The characteristic information of the image includes structural joints, cracks, water seepage points, etc.

[0065] S2: Evenly divide the two-dimensional image of the tunnel inner wall into several square areas and establish a coordinate system. In this embodiment, the high-precision two-dimensional image obtained by scanning the tunnel inner wall using machine vision technology is evenly divided into several square areas and a coordinate system is established. Figure 2 As shown. The side length of the square area is , which is equal to the diameter of the anchor hole. The hollow circle indicates the position of the anchor hole, and the black curve indicates the cracks on the inner wall of the tunnel.

[0066] S3: Based on the square area with the coordinate system, determine whether the preset anchor holes of the tunnel contact network avoid cracks, water seepage points and other unfavorable conditions. If so, determine the current anchor hole position; if not, use the gray wolf optimization algorithm to randomly generate artificial wolves in the square area with the coordinate system; where each artificial wolf position represents an anchor hole position, as shown in the attached figure. Figure 2 Shown as grey diamond.

[0067] S4: Establish a fitness function based on the preset objective function and the current position of the artificial wolf.

[0068] A further implementation is that the method for obtaining the fitness function includes:

[0069] Based on the coordinate system, the position coordinates of the artificial wolf are obtained, and based on the position coordinates, the position of the artificial wolf is obtained. The calculation formula for the position of the artificial wolf is as follows:

[0070]

[0071] in, Indicates The location coordinates of the artificial wolf , and are the upper and lower bounds of the spatial search range, and the position of each artificial wolf .

[0072] Based on the position and weight coefficient of the artificial wolf, a fitness function is obtained; wherein the weight coefficient includes a first weight coefficient for adjusting the influence ratio of the benefit in the fitness, a second weight coefficient for adjusting the proportion of the energy loss in the fitness, and a third weight coefficient for adjusting the proportion of the risk cost; the expression of the fitness function is as follows:

[0073]

[0074] in, Represents the objective function value, measuring the current position Optimization effect; is the first weight coefficient, Indicates the current position of the artificial wolf With preset position distance; The second weight coefficient, is represented as a set of danger zones, is the indicator function, if The value is 1 in the danger zone and 0 otherwise. is the third weight coefficient.

[0075] S5: Sort the wolf pack based on the fitness function to obtain the alpha wolf, scout wolf and fierce wolf; specifically, according to the fitness value Function. Sort the wolf pack and select three wolves: the alpha wolf (leader, with the highest fitness value), the scout wolf (co-leader, with the second highest fitness value), and the fierce wolf (explorer, with the third highest fitness value).

[0076] S6: Construct a feedforward neural network, combine the intelligent behaviors of the alpha wolf, scout wolf and fierce wolf, and obtain the position of the alpha wolf and the fitness function value of the alpha wolf; among them, the position of the alpha wolf is the optimal position of the current anchor hole, completing the positioning of the anchor hole of the tunnel contact network.

[0077] A further implementation method is that the intelligent behavior includes the rules for the generation of alpha wolf, the wandering behavior of scout wolves, and the behavior of the alpha wolf summoning fierce wolves.

[0078] A further implementation method is that the method for obtaining the position of the alpha wolf and the fitness function value of the alpha wolf includes:

[0079] An artificial wolf with an objective function value that meets the preset optimal requirements is selected as the alpha wolf; specifically, in the initial solution space, the objective function value is calculated to find the individual with the optimal objective function value as the alpha wolf. In the process of the cycle, the value of the objective function of the best wolf after each cycle is compared with the value of the leader in the previous generation. If it is better than the previous one, the position of the leader is adjusted. If there is more than one, one of them is randomly selected as the new leader. The new leader no longer participates in the three intelligent actions, but immediately starts the next cycle until it is replaced by a more powerful artificial wolf.

[0080] The direction in which the objective function value is greater than the objective function value of the current position is taken as the wandering direction; based on the wandering direction and the preset wandering step length, the scout wolf repeatedly performs the wandering behavior until the objective function value perceived by the scout wolf is greater than the objective function value of the current alpha wolf or the scout wolf reaches the preset wandering times, then the scout wolf is updated to the alpha wolf and a summoning behavior is initiated; specifically, the scout wolf searches for prey. The best performing artificial wolf in the solution space except the alpha wolf is regarded as the scout wolf, and then the prey is searched in the solution space, and one is randomly selected as the integer scout wolf's proportional factor α. Select the one with the strongest smell (largest objective function value) and greater than the current position odor concentration One step forward in the direction of. As shown in formula (3):

[0081]

[0082] in, Indicates the updated position coordinates of Tanlang. Indicates Only explore the wolf's current location coordinates. is the walking step length, h is the walking direction, Repeat the above wandering behavior until there is a concentration of scent of a certain wolf. , at this time, the scout wolf replaces the alpha wolf and initiates the call behavior, or reaches the number of wandering times .

[0083] When the alpha wolf summons the wolves, the alpha wolf guides the wolves to charge toward the alpha wolf's current position with a preset charging step length, and obtains a feature vector based on the obstacle density, spatial scale, and distance between the wolves and the alpha wolf in the charging path.

[0084] The feature vector is used as the input vector of the feedforward neural network to obtain the predicted running step length; specifically, when the alpha wolf is updated, it means that a higher objective function value has been found. At this moment, the alpha wolf can start commanding the task and guide the fierce wolves to quickly move towards the place where the alpha wolf is now with a step length of Run and attack, as shown in formula (4):

[0085]

[0086] in is the position of the k-th generation group leader in space, It is the running stride length. Indicates The current location of the wolf. Indicates the updated position of the wolf. The wolf does not just rush forward, but senses the concentration of the prey's scent around it during the rush.

[0087] Obstacle density in the running paths of fierce wolves and alpha wolves , space scale , the distance between the fierce wolf and the leader wolf, as shown in formula (5):

[0088]

[0089] These input features are normalized to form the input vector of the neural network, as shown in formula (6):

[0090]

[0091] The fierce wolf attacks based on the predicted attack step length. When the objective function value perceived by the fierce wolf is greater than the objective function value of the current alpha wolf or the fierce wolf reaches the preset attack condition, the fierce wolf is updated to the alpha wolf, and the updated alpha wolf's position and the alpha wolf's fitness function are obtained.

[0092] In a further embodiment, the feedforward neural network comprises an input layer, a hidden layer and an output layer.

[0093] The input layer includes three nodes, corresponding to the obstacle density, spatial scale, and the distance between the wolf and the leader, that is, the corresponding features .

[0094] The hidden layer consists of 2 layers, each layer contains 4 neurons, and the activation function is , if you enter If it is greater than 0, then output itself, otherwise the output is 0;

[0095] The output layer includes 1 node, which is used to output the predicted running step length. Target step length With prediction step length The mean square error (MSE) is shown in formula (7):

[0096]

[0097] Since the neural network is used to dynamically output the running step length, each time the wolf runs, according to the current state Predicting the appropriate step size using a trained neural network , and update the position as shown in formula (8):

[0098]

[0099] The wolf sensed the concentration of the prey's scent while running. , triggering the alpha wolf update mechanism, thus , making the wolf the new generation leader, replacing the original leader to command. If the wolf does not sense that the concentration of the prey's odor is greater than that of the current leader during the charge, it will continue to charge until the maximum number of iterations is reached.

[0100] When the leader reaches the optimization accuracy requirement or the algorithm reaches the maximum number of iterations, the leader's position and the output value of the leader's fitness function are output, and the leader's position is used as the optimal position of the anchor hole.

[0101] In summary, according to the fitness function Evaluate the starting point of the artificial wolves and select the artificial wolf with the largest fitness function value as the leader The other artificial wolves are regarded as scout wolves, responsible for implementing wandering behavior until one of the scout wolves perceives the target function The value exceeds the objective function value of the alpha wolf , or the maximum number of walks set is reached . Then the alpha wolf performs a summoning action, guiding the wolves to charge in the direction of the alpha wolf. As a dynamic adjustment parameter, a neural network is used for training so that it can be adaptively adjusted according to environmental characteristics and wolf pack status.

[0102] Embodiment 2

[0103] The present invention also provides an anchor hole positioning system based on a neural network and a gray wolf optimization algorithm, which is used to implement a method, including:

[0104] A two-dimensional image acquisition module is used to scan the inner wall of the tunnel based on machine vision technology to obtain a two-dimensional image of the inner wall of the tunnel;

[0105] An image segmentation module is used to evenly segment the two-dimensional image of the tunnel inner wall to obtain a number of square areas and establish a coordinate system;

[0106] The artificial wolf generation module is used to determine whether the preset anchor holes of the tunnel contact network avoid cracks, water seepage points and other unfavorable conditions based on the square area with the coordinate system. If so, the current anchor hole position is determined; if not, the gray wolf optimization algorithm is used to randomly generate artificial wolves in the square area with the coordinate system; wherein each artificial wolf position represents an anchor hole position;

[0107] A fitness function building module is used to build a fitness function based on a preset objective function and the current position of the artificial wolf;

[0108] The wolf pack sorting module is used to sort the wolf pack based on the fitness function to obtain the alpha wolf, scout wolf and fierce wolf;

[0109] The optimal anchor hole positioning module is used to construct a feedforward neural network, combining the intelligent behaviors of the alpha wolf, scout wolf and fierce wolf to obtain the position of the alpha wolf and the fitness function value of the alpha wolf; among them, the position of the alpha wolf is the optimal position of the current anchor hole, completing the positioning of the anchor hole of the tunnel contact network.

[0110] A further implementation is that the fitness function building module includes:

[0111] The artificial wolf position acquisition unit is used to obtain the position coordinates of the artificial wolf based on the coordinate system, and obtain the position of the artificial wolf based on the position coordinates. The calculation formula of the artificial wolf position is as follows:

[0112] ,

[0113] Among them, among them, Indicates The location coordinates of the artificial wolf , and are the upper and lower bounds of the spatial search range, and the position of each artificial wolf ;

[0114] The fitness function calculation unit is used to obtain the fitness function based on the position and weight coefficient of the artificial wolf; wherein the weight coefficient includes a first weight coefficient for adjusting the influence ratio of the benefit in the fitness, a second weight coefficient for adjusting the proportion of the energy loss in the fitness, and a third weight coefficient for adjusting the proportion of the risk cost; the fitness function is expressed as follows:

[0115] ,

[0116] in, Represents the objective function value, measuring the current position Optimization effect; is the first weight coefficient, Indicates the current position of the artificial wolf With preset position distance; The second weight coefficient, is represented as a set of danger zones, is the indicator function, is the third weight coefficient.

[0117] A further implementation method is that the intelligent behavior includes the rules for the generation of alpha wolf, the wandering behavior of scout wolves, and the behavior of the alpha wolf summoning fierce wolves.

[0118] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. An anchor hole positioning method based on neural network and gray wolf optimization algorithm, characterized in that: The following steps are involved: Scan the inner wall of the tunnel based on machine vision technology to obtain a two-dimensional image of the inner wall of the tunnel; Evenly dividing the two-dimensional image of the inner wall of the tunnel to obtain a number of square areas, and establishing a coordinate system; Based on the square area with the coordinate system, it is determined whether the preset anchor holes of the tunnel contact network avoid cracks and water seepage points. If so, the current anchor hole position is determined; if not, the gray wolf optimization algorithm is used to randomly generate artificial wolves in the square area with the coordinate system; wherein each artificial wolf position represents an anchor hole position; Based on the preset objective function and the current position of the artificial wolf, a fitness function is established; Sorting the wolf pack based on the fitness function to obtain the alpha wolf, the scout wolf and the fierce wolf; A feedforward neural network is constructed, and the intelligent behaviors of the alpha wolf, scout wolf and fierce wolf are combined to obtain the position of the alpha wolf and the fitness function value of the alpha wolf; wherein the position of the alpha wolf is the optimal position of the current anchor hole, and the positioning of the anchor hole of the tunnel contact network is completed.

2. The method according to claim 1, characterized in that The method for obtaining the fitness function includes: Based on the coordinate system, the position coordinates of the artificial wolf are obtained, and based on the position coordinates, the position of the artificial wolf is obtained. The calculation formula of the position of the artificial wolf is as follows: ， in, Indicates The location coordinates of the artificial wolf , and are the upper and lower bounds of the spatial search range, and the position of each artificial wolf ; Based on the position and weight coefficient of the artificial wolf, the fitness function is obtained; wherein the weight coefficient includes a first weight coefficient for adjusting the influence ratio of the benefit in the fitness, a second weight coefficient for adjusting the proportion of the energy loss in the fitness, and a third weight coefficient for adjusting the proportion of the risk cost; the fitness function is expressed as follows: ， in, Represents the objective function value, measuring the current position Optimization effect; is the first weight coefficient, Indicates the current position of the artificial wolf With preset position distance; The second weight coefficient, is represented as a set of danger zones, is the indicator function, is the third weight coefficient.

3. The method according to claim 2, characterized in that The intelligent behaviors include the rules generated by the alpha wolf, the wandering behaviors of scout wolves, and the behavior of the alpha wolf summoning fierce wolves.

4. The method according to claim 3, characterized in that Methods for obtaining the position of the alpha wolf and the fitness function value of the alpha wolf include: An artificial wolf with an objective function value that meets the preset optimal requirements is selected as the leader wolf; The direction in which the objective function value is greater than the objective function value of the current position is taken as the wandering direction; based on the wandering direction and the preset wandering step length, the scout wolf repeatedly performs the wandering behavior until the objective function value perceived by the scout wolf is greater than the objective function value of the current alpha wolf or the scout wolf reaches a preset wandering number, then the scout wolf is updated to the alpha wolf and a summoning behavior is initiated; In the behavior of the alpha wolf calling the fierce wolf, the alpha wolf guides the fierce wolf to rush towards the alpha wolf's current position with a preset rushing step length, and obtains a feature vector based on the obstacle density, spatial scale and distance between the fierce wolf and the alpha wolf in the rushing path of the fierce wolf and the alpha wolf; Using the feature vector as an input vector of the feedforward neural network to obtain a predicted running step length; The fierce wolf charges based on the predicted charging step length. When the objective function value perceived by the fierce wolf is greater than the objective function value of the current alpha wolf or the fierce wolf reaches a preset charging condition, the fierce wolf is updated to the alpha wolf, and the updated position of the alpha wolf and the fitness function of the alpha wolf are obtained.

5. The method according to claim 4, characterized in that The feedforward neural network includes an input layer, a hidden layer and an output layer; The input layer includes three nodes, corresponding to the obstacle density, the spatial scale, and the distance between the fierce wolf and the leader wolf respectively; The hidden layer consists of 2 layers, each containing 4 neurons; The output layer includes 1 node, which is used to output the predicted running step length.

6. An anchor hole positioning system based on a neural network and a gray wolf optimization algorithm, used to implement the method described in any one of claims 1 to 5, characterized in that: include: A two-dimensional image acquisition module is used to scan the inner wall of the tunnel based on machine vision technology to obtain a two-dimensional image of the inner wall of the tunnel; An image segmentation module is used to evenly segment the two-dimensional image of the inner wall of the tunnel to obtain a number of square areas and establish a coordinate system; The artificial wolf generation module is used to determine whether the preset anchor holes of the tunnel contact network avoid cracks and water seepage points based on the square area with the coordinate system. If so, the current anchor hole position is determined; if not, the gray wolf optimization algorithm is used to randomly generate artificial wolves in the square area with the coordinate system; wherein each artificial wolf position represents an anchor hole position; A fitness function building module is used to build a fitness function based on a preset objective function and the current position of the artificial wolf; A wolf pack sorting module is used to sort the wolf pack based on the fitness function to obtain the alpha wolf, the scout wolf and the fierce wolf; The optimal anchor hole positioning module is used to construct a feedforward neural network, combining the intelligent behaviors of the alpha wolf, the scout wolf and the fierce wolf to obtain the position of the alpha wolf and the fitness function value of the alpha wolf; wherein the position of the alpha wolf is the optimal position of the current anchor hole, completing the positioning of the anchor hole of the tunnel contact network.

7. The system according to claim 6, characterized in that The fitness function building module includes: The artificial wolf position acquisition unit is used to obtain the position coordinates of the artificial wolf based on the coordinate system, and obtain the position of the artificial wolf based on the position coordinates. The calculation formula of the artificial wolf position is as follows: ， in, Indicates The location coordinates of the artificial wolf , and are the upper and lower bounds of the spatial search range, and the position of each artificial wolf ; The fitness function calculation unit is used to obtain the fitness function based on the position and weight coefficient of the artificial wolf; wherein the weight coefficient includes a first weight coefficient for adjusting the influence ratio of the benefit in the fitness, a second weight coefficient for adjusting the proportion of the energy loss in the fitness, and a third weight coefficient for adjusting the proportion of the risk cost; the fitness function is expressed as follows: ， in, Represents the objective function value, measuring the current position Optimization effect; is the first weight coefficient, Indicates the current position of the artificial wolf With preset position distance; The second weight coefficient, is represented as a set of danger zones, is the indicator function, is the third weight coefficient.

8. The system according to claim 6, characterized in that The intelligent behaviors include the rules generated by the alpha wolf, the wandering behaviors of scout wolves, and the behavior of the alpha wolf summoning fierce wolves.

Images