A method, device, equipment and medium for obtaining engineering control points along a line
Through visual state space model and deep learning methods, the problem of omission of control elements in railway engineering surveys is solved, and fast and accurate data extraction is achieved.
Patent Information
- Application Number
- CN202411250313.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-06
- Publication Date
- 2025-07-08
- Estimated Expiration
- 2044-09-06
AI Technical Summary
In railway engineering survey, due to the long survey lines and complex or changeable terrain, designers are prone to miss important control elements during data processing and cannot quickly and accurately extract data.
By acquiring the visual state space model, multi-dimensional feature image extraction and processing are performed, combined with decoder, network deep supervision method and loss function optimization, the landing model along the line is obtained, and control points are obtained using the overlay analysis method.
It realizes rapid and accurate acquisition of engineering control points along the line under complex terrain conditions, avoids the omission of important elements, and improves the accuracy and efficiency of data extraction.
Smart Images
Figure CN119418157B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of railway engineering, and in particular, to a method, device, equipment and medium for obtaining engineering control points along a line. Background Art
[0002] In the existing railway engineering technology, survey and design personnel conduct surveys on roads, existing railways, rivers, buildings, etc. Due to factors such as long survey lines, complex or variable terrain and ground features, it is easy for designers to miss some important control elements during the process of processing survey data, and it is impossible to ensure that data can be extracted quickly and accurately. Therefore, there is an urgent need for a method, device, equipment and medium for obtaining engineering control points along a line to solve the problem that due to factors such as long survey lines, complex or variable terrain and ground features, it is easy for designers to miss some important control elements during the process of processing survey data, and it is impossible to ensure that data can be extracted quickly and accurately. Summary of the Invention
[0003] The purpose of the present invention is to provide a method, device, equipment and medium for obtaining engineering control points along a line to improve the above problems. To achieve the above purpose, the technical solutions adopted by the present invention are as follows:
[0004] In the first aspect, the present application provides a method for obtaining engineering control points along a line, including:
[0005] Obtain a visual state space model;
[0006] Based on the visual state space model, perform image extraction on preset multi-dimensional features to obtain image features of parallel branches;
[0007] Perform image processing on the image features of the parallel branches to obtain a feature map of the parallel branches;
[0008] Based on a preset decoder, perform feature extraction and fusion on the feature map of the parallel branches to obtain multi-scale features of the image;
[0009] According to a preset network deep supervision method, perform feature extraction on the image features of the parallel branches and the multi-scale features of the image to obtain a ground object model along the line;
[0010] Based on a preset loss function and a preset network deep supervision method, perform optimization and solution on the ground object model along the line to obtain an optimal ground object model along the line;
[0011] According to a preset overlay analysis method, perform solution on the optimal ground object model along the line to obtain engineering control points along the line.
[0012] In the second aspect, the present application also provides a device for obtaining engineering control points along a line, including:
[0013] A first acquisition module, configured to acquire a visual state space model;
[0014] A first processing module, configured to perform image extraction on preset multi-dimensional features based on the visual state space model to obtain image features of parallel branches;
[0015] A second processing module, configured to perform image processing on the image features of the parallel branches to obtain a feature map of the parallel branches;
[0016] A third processing module, configured to perform feature extraction and fusion on the feature map of the parallel branches based on a preset decoder to obtain multi-scale features of an image;
[0017] A fourth processing module, configured to perform feature extraction on the image features of the parallel branches and the multi-scale features of the image according to a preset network deep supervision method to obtain a ground object model along the line;
[0018] A fifth processing module, configured to perform optimization and solution on the ground object model along the line based on a preset loss function and a preset network deep supervision method to obtain an optimal ground object model along the line;
[0019] A sixth processing module, configured to perform solution on the optimal ground object model along the line according to a preset overlay analysis method to obtain control points of a project along the line.
[0020] In a third aspect, the present application further provides a device for acquiring control points of a project along the line, including:
[0021] A memory, configured to store a computer program;
[0022] A processor, configured to implement the steps of the method for acquiring control points of a project along the line when executing the computer program.
[0023] In a fourth aspect, the present application further provides a readable storage medium, on which a computer program is stored, and the computer program, when executed by a processor, implements the steps of the above-mentioned method for acquiring control points of a project along the line.
[0024] The beneficial effects of the present invention are as follows:
[0025] The present invention obtains a ground object model along the line through the image features of the parallel branches and the multi-scale features of the image, performs optimization and solution on the ground object model along the line according to a loss function and a network deep supervision method, and then performs processing through an overlay analysis method to obtain control points of a project along the line. It solves the problems that due to factors such as a long survey line, complex or variable terrain and ground objects, designers are prone to miss some important control elements during the process of processing survey data, and it is impossible to ensure that data can be quickly and accurately extracted.
[0026] Other features and advantages of the present invention will be described in the following specification, and in part will be obvious from the specification, or can be understood by implementing the embodiments of the present invention. The objectives and other advantages of the present invention can be achieved and obtained by the structures specifically pointed out in the written specification, claims, and drawings. Brief Description of the Drawings
[0027] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention, and thus should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.
[0028] Figure 1 It is a schematic flow chart of the method for obtaining engineering control points along the line described in the embodiments of the present invention;
[0029] Figure 2 It is a schematic structural diagram of the device for obtaining engineering control points along the line described in the embodiments of the present invention.
[0030] Reference numerals in the figure: 800, device for obtaining engineering control points along the line; 801, processor; 802, memory; 803, multimedia component; 804, I / O interface; 805, communication component. Detailed Embodiments
[0031] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Usually, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed present invention, but merely represents the selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.
[0032] It should be noted that: similar reference numerals and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of the present invention, terms such as "first", "second", etc. are only used for distinguishing descriptions and cannot be understood as indicating or implying relative importance.
[0033] Embodiment 1:
[0034] This embodiment provides a method for obtaining engineering control points along a line.
[0035] Refer to Figure 1 , the figure shows that this method includes steps S1 to S7, including:
[0036] S1: Obtain a visual state space model;
[0037] S2: Based on the visual state space model, perform image extraction on preset multi-dimensional features to obtain image features of parallel branches;
[0038] The visual state space model includes a feature selection module.
[0039] To clarify the specific acquisition method of the image features of parallel branches, steps S21 to S25 are included in step S2, specifically:
[0040] S21: Obtain images along the engineering line;
[0041] S22: Perform geometric correction and radiometric correction on the images along the engineering line to obtain a target image;
[0042] In this step, the geometric correction is a process for eliminating or correcting geometric errors of remote sensing images, and the geometric correction includes geometric rough correction, geometric fine correction, hybrid correction, and orthorectification; the radiometric correction is a correction for systematic and random radiometric distortions or aberrations generated by the data acquisition and transmission system, and is a process for eliminating or correcting the image brightness aberration caused by radiometric errors.
[0043] S23: Based on a preset ground object extraction network model, perform convolution and pooling operations on the target image to obtain initial features of the image;
[0044] In this step, the target image first passes through a 3×3 convolution with 32 convolution kernels and a stride of 2; the target image then passes through a 3×3 convolution with 32 convolution kernels and a stride of 1; the target image finally passes through a 3×3 convolution with 96 convolution kernels and a stride of 1 and a 3×3 max pooling with a stride of 2 to obtain initial features of the image;
[0045] The initial features of the image are:
[0046]
[0047] In the above formula (1): H represents the height of the target image, W represents the width of the target image, and 96 represents the number of convolution kernels;
[0048] Preferably, the feature extraction network model is ResNet Stem, and the Stem in ResNet Stem is to reduce the size of the input image, extract the preliminary features of the input image, reduce the computational load of subsequent layers, improve network efficiency and stabilize the learning process.
[0049] S24: Perform multi-dimensional feature extraction on the initial features of the image to obtain multi-dimensional features;
[0050] The feature extraction network model includes a deep residual network model;
[0051] To clarify the specific acquisition method of the multi-dimensional features, steps S241 to S247 are included in step S24, specifically:
[0052] S241: Solve the initial features of the image according to the deep residual network model to obtain an initial feature map;
[0053] In this step, input the initial features of the image into the deep residual network model, and solve the input deep residual network model through the backbone part of ResNet in the ResNet Stem to obtain an initial feature map.
[0054] S242: Perform deformable convolution processing on the initial feature map to obtain the first branch feature;
[0055] S243: Process the first branch feature based on the feature selection module to obtain the processed first branch feature;
[0056] Preferably, the feature selection module is two consecutive VSS modules.
[0057] S244: Perform deformable convolution processing on the processed first branch feature to obtain the second branch feature;
[0058] S245: Perform integrated depth feature processing on the second branch feature to obtain the integrated second branch feature;
[0059] In this step, the integrated depth feature is the fusion of upsampling and the first branch feature. Perform upsampling and first branch feature fusion processing on the second branch feature to obtain the processed second branch feature, and the feature selection module processes the processed second branch feature to obtain the integrated second branch feature;
[0060] Preferably, the feature selection module is two consecutive VSS modules.
[0061] S246: Perform deformable convolution processing on the first branch feature and the integrated second branch feature to obtain the third branch feature;
[0062] In this step, the fourth branch feature is further included. Based on the feature selection module, the first branch feature, the second branch feature, and the third branch feature are processed to obtain the fourth branch feature;
[0063] Preferably, the feature selection module is nine consecutive VSS modules.
[0064] S247: Take the first branch feature, the second branch feature, and the third branch feature as multi-dimensional features.
[0065] S25: Based on the feature selection module, image extraction is performed on the multi-dimensional features to obtain the image features of the parallel branches.
[0066] Preferably, the feature selection module is two consecutive VSS modules.
[0067] S3: Perform image processing on the image features of the parallel branches to obtain the feature map of the parallel branches;
[0068] S4: Based on a preset decoder, feature extraction and fusion are performed on the feature map of the parallel branches to obtain the multi-scale features of the image;
[0069] The decoder includes an attention module;
[0070] To clarify the specific acquisition method of the multi-scale features of the image, steps S41 to S45 are included in step S4, specifically:
[0071] S41: According to the preset separation ratio in the attention module, feature extraction and fusion are performed on the feature map of the parallel branches to obtain the spatial feature of the feature map and the channel feature of the feature map;
[0072] In this step, the separation ratio is two groups of feature maps that divide the channels. The two groups of feature maps that divide the channels perform feature extraction and fusion on the feature map of the parallel branches to obtain the spatial feature of the feature map and the channel feature of the feature map.
[0073] S42: Perform convolution adjustment on the preset number of channels of the feature map of the parallel branches to obtain the convolution channel number;
[0074] In this step, the convolution adjustment is 1×1 convolution adjustment.
[0075] S43: Solve the feature map of the parallel branches according to a preset logical function to obtain the attention feature of the feature map;
[0076] The logical function includes a normalization function and an S-shaped growth curve function;
[0077] To clarify the specific acquisition method of the attention features of the feature map, steps S43 includes S431 to S436, specifically:
[0078] S431: Perform global average pooling on the feature map of the parallel branch to obtain the average pooling feature;
[0079] S432: Perform global max pooling on the feature map of the parallel branch to obtain the max pooling feature;
[0080] S433: Construct based on the average pooling feature and the max pooling feature to obtain the hybrid pooling feature;
[0081] Preferably, the average pooling feature and the max pooling feature are added to obtain the hybrid pooling feature.
[0082] S434: Perform hybrid pooling feature processing on the feature map of the parallel branch according to the normalization function to obtain the channel attention magnitude;
[0083] Preferably, the channel attention magnitude is a weight within the range of 0 - 1.
[0084] S435: Perform global average pooling on the feature map of the parallel branch according to the S-shaped growth curve function to obtain the spatial attention weight;
[0085] In this step, after performing global average pooling on the feature map of the parallel branch in the x-direction and y-direction respectively for spatial attention, it is processed through the S-shaped growth curve function to obtain the spatial attention weight.
[0086] S436: Construct based on the channel attention magnitude and the spatial attention weight to obtain the attention features of the feature map.
[0087] S44: Calculate different convolution kernels and different dilation rates on the feature map of the parallel branch based on the attention module to obtain a convolution data set;
[0088] To clarify the specific acquisition method of the convolution data set, steps S44 includes S441 to S444, specifically:
[0089] S441: Input the feature map of the parallel branch into the attention module, and perform upsampling on the feature map of the parallel branch through the bilinear interpolation method to obtain the operation results of the first branch and the second branch;
[0090] S442: Calculate different convolution kernels and different dilation rates on the operation result of the first branch to obtain the first convolution data;
[0091] Preferably, the different convolution kernels and different dilation rates are calculated as convolution branches with different convolution kernels and different dilation rates; the convolution branches with different convolution kernels and different dilation rates include: convolution kernel 5 and dilation rate 1; convolution kernel 7 and dilation rate 1; convolution kernel 3 and dilation rate 6; convolution kernel 3 and dilation rate 12; convolution kernel 3 and dilation rate 18.
[0092] S443: Calculate the different convolution kernels and different dilation rates for the operation result of the second branch to obtain second convolution data;
[0093] S444: Construct a convolution data set according to the first convolution data and the second convolution data.
[0094] In this step, all different convolution kernels and different dilation rates are added and then the first convolution data and the second convolution data are calculated to obtain a convolution data set.
[0095] S45: Based on the spatial features of the feature map, the channel features of the feature map, the number of convolution channels, the attention features of the feature map, and the convolution data set, construct multi-scale features of the image.
[0096] In this step, construction is performed through upsampling.
[0097] S5: According to a preset network deep supervision method, extract features from the image features of the parallel branches and the multi-scale features of the image to obtain a linear object model;
[0098] S6: Based on a preset loss function and a preset network deep supervision method, optimize and solve the linear object model to obtain an optimal linear object model;
[0099] The loss function includes an edge loss function and a cross-entropy loss function;
[0100] To clarify the specific acquisition method of the optimal linear object model, steps S61 to S65 are included in step S6, specifically:
[0101] S61: Obtain the label contour information of the target image, and the label contour information of the target image includes true label contour values, the number of first internal pixels, and the number of second internal pixels;
[0102] S62: Calculate the true label contour values based on a preset loss function to obtain a loss value;
[0103] To clarify the specific acquisition method of the loss value, steps S621 to S623 are included in step S62, specifically:
[0104] S621: Calculate the total number of pixels based on the first internal pixel quantity and the second internal pixel quantity;
[0105] In this step, the total number of pixels is:
[0106] N = |X| + |Y| (2)
[0107] In the above formula (2): N represents the total number of pixels, |X| represents the number of internal pixels in X, and |Y| represents the number of internal pixels in Y;
[0108] The number of internal pixels in X represents the first internal pixel quantity, and the number of internal pixels in Y represents the second internal pixel quantity.
[0109] S622: Calculate the edge loss value of the sample pixels based on the edge loss function for the true label contour value and the predicted label contour value;
[0110] In this step, the edge loss value of the sample pixels is:
[0111]
[0112] In the above formula (3): L edge represents the edge loss value of the sample pixels, N represents the total number of pixels, x i represents the predicted label contour value, and y i represents the true label contour value.
[0113] S623: Calculate the cross-entropy loss value between the output result and the true label based on the cross-entropy loss function for the total number of pixels and the predicted pixel probabilities;
[0114] In this step, the cross-entropy loss value between the output result and the true label is:
[0115]
[0116] In the above formula (4): CELoss represents the cross-entropy loss value between the output result and the true label, N represents the total number of pixels, M represents the number of categories, is the label variable in one-hot form, represents the probability that the predicted pixel i belongs to category c;
[0117] The probability that the predicted pixel i belongs to category c represents the predicted pixel probability.
[0118] S63: Calculate the sample result based on the preset network deep supervision method for the first internal pixel quantity and the second internal pixel quantity;
[0119] In this step, the sample result is as follows:
[0120]
[0121] In the above formula (5): Loss dice (X, Y) represents the sample result of (X, Y) under the deep supervision of the network; |X∩Y| represents the number of pixels in the intersection of X and Y, |X| represents the number of pixels inside X, |Y| represents the number of pixels inside Y, X represents the result of sample prediction, and Y represents the true label of the dataset;
[0122] The sample result under the deep supervision of the network represents the sample result, the number of pixels inside X represents the first number of internal pixels, and the number of pixels inside Y represents the second number of internal pixels.
[0123] S64: Construct according to the loss value, the sample result, the preset weight parameter, and the regularization term of the preset weight parameter to obtain the calculation formula of the ground object loss;
[0124] In this step, the calculation formula of the ground object loss is as follows:
[0125]
[0126] In the above formula (6): Loss represents the ground object loss function, α1, α2, and α3 all represent learnable weight parameters, Loss dice represents the sample result under the deep supervision of the network, CELoss represents the cross-entropy loss value between the output result and the true label, L edge represents the edge loss value of the sample pixels, represents the L2 regularization term calculated by the weight parameter;
[0127] The learnable weight parameter represents the weight parameter, and the L2 regularization term calculated by the weight parameter represents the regularization term of the preset weight parameter.
[0128] S65: Optimize and solve the ground object model along the line based on the calculation formula of the ground object loss to obtain the optimal ground object model along the line.
[0129] S7: Solve the optimal ground object model along the line according to the preset overlay analysis method to obtain the control points of the project along the line.
[0130] To clarify the specific acquisition method of the control points of the project along the line, steps S7 includes S71 to S74, specifically:
[0131] S71: Predict the remote sensing image of the optimal ground object model along the line to obtain the first ground object category and the second ground object category;
[0132] S72: Construct according to the first type of ground object and the second type of ground object to obtain the ground object classification result;
[0133] In this step, merge the first type of ground object and the second type of ground object to obtain the ground object classification result.
[0134] S73: Solve the ground object classification result based on a preset grid space data conversion library and a preset image processing library to obtain the ground object boundary;
[0135] In this step, convert the ground object classification result into vector data, set a threshold based on the grid space data conversion library to delete the fine parts of the vector data, and remove the holes in the vector data based on the image processing library to obtain the ground object boundary.
[0136] S74: Solve the ground object boundary and preset railway data based on a preset overlay analysis method to obtain the control points of the project along the line.
[0137] In this step, perform vector overlay analysis on the vector data and the railway data to obtain the control points of the project along the line.
[0138] Embodiment 2:
[0139] This embodiment provides a device for obtaining the control points of the project along the line, and the device includes:
[0140] The first acquisition module is used to acquire the visual state space model;
[0141] The first processing module is used to perform image extraction on preset multi-dimensional features based on the visual state space model to obtain the image features of the parallel branches;
[0142] To clarify the specific acquisition method of the image features of the parallel branches, specifically:
[0143] The first acquisition unit is used to acquire the images along the project;
[0144] The first correction unit is used to perform geometric correction and radiometric correction on the images along the project to obtain the target image;
[0145] The first processing unit is used to perform convolution and pooling operations on the target image based on a preset ground object extraction network model to obtain the initial features of the image;
[0146] The second processing unit is used to perform multi-dimensional feature extraction on the initial features of the image to obtain multi-dimensional features;
[0147] The third processing unit is used to perform image extraction on the multi-dimensional features based on the feature selection module to obtain the image features of the parallel branches.
[0148] A second processing module for performing image processing on the image features of the parallel branches to obtain a feature map of the parallel branches;
[0149] A third processing module for performing feature extraction and fusion on the feature map of the parallel branches based on a preset decoder to obtain multi-scale features of the image;
[0150] A fourth processing module for performing feature extraction on the image features of the parallel branches and the multi-scale features of the image according to a preset network deep supervision method to obtain a ground object model along the line;
[0151] A fifth processing module for optimizing and solving the ground object model along the line based on a preset loss function and a preset network deep supervision method to obtain an optimal ground object model along the line;
[0152] A sixth processing module for solving the optimal ground object model along the line according to a preset overlay analysis method to obtain control points of the project along the line.
[0153] To clarify the specific acquisition method of the image features of the parallel branches, specifically:
[0154] A first construction unit for predicting remote sensing images of the optimal ground object model along the line to obtain a first ground object category and a second ground object category;
[0155] A second construction unit for constructing according to the first ground object category and the second ground object category to obtain a ground object classification result;
[0156] A third construction unit for solving the ground object classification result based on a preset raster spatial data conversion library and a preset image processing library to obtain a ground object boundary;
[0157] A fourth construction unit for solving the ground object boundary and preset railway data based on a preset overlay analysis method to obtain control points of the project along the line.
[0158] It should be noted that regarding the device in the above embodiments, the specific manner in which each module performs operations has been described in detail in the embodiments related to the method, and will not be elaborated here.
[0159] Embodiment 3:
[0160] Corresponding to the above method embodiment, in this embodiment, an acquisition device for control points of a project along the line is also provided. The acquisition device for control points of a project along the line described below can be mutually corresponding and referred to with the acquisition method for control points of a project along the line described above.
[0161] Figure 2It is a block diagram of an acquisition device 800 for engineering control points along a line shown according to an exemplary embodiment. As Figure 2 shown, the acquisition device 800 for engineering control points along the line may include: a processor 801, a memory 802. The acquisition device 800 for engineering control points along the line may also include one or more of a multimedia component 803, an I / O interface 804, and a communication component 805.
[0162] Among them, the processor 801 is used to control the overall operation of the acquisition device 800 for engineering control points along the line to complete all or part of the steps in the above-mentioned method for acquiring engineering control points along the line. The memory 802 is used to store various types of data to support the operation of the acquisition device 800 for engineering control points along the line. These data may include, for example, instructions for any application or method operating on the acquisition device 800 for engineering control points along the line, as well as application-related data, such as contact data, sent and received messages, pictures, audio, video, and so on. The memory 802 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic memory, flash memory, a magnetic disk, or an optical disc. The multimedia component 803 may include a screen and an audio component. The screen may be a touch screen, for example, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signal may be further stored in the memory 802 or sent through the communication component 805. The audio component further includes at least one speaker for outputting audio signals. The I / O interface 804 provides an interface between the processor 801 and other interface modules, and the above-mentioned other interface modules may be a keyboard, a mouse, buttons, etc. These buttons may be virtual buttons or physical buttons. The communication component 805 is used for wired or wireless communication between the acquisition device 800 for engineering control points along the line and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination of one or more of them. Accordingly, the communication component 805 may include: a Wi-Fi module, a Bluetooth module, and an NFC module.
[0163] In an exemplary embodiment, the acquisition device 800 for engineering control points along a line may be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components, and is used to execute the above-mentioned method for acquiring engineering control points along a line.
[0164] In another exemplary embodiment, a computer-readable storage medium including program instructions is further provided. When the program instructions are executed by a processor, the steps of the above-mentioned method for acquiring engineering control points along a line are implemented. For example, the computer-readable storage medium may be the above-mentioned memory 802 including program instructions, and the above-mentioned program instructions may be executed by the processor 801 of the acquisition device 800 for engineering control points along a line to complete the above-mentioned method for acquiring engineering control points along a line.
[0165] Embodiment 4:
[0166] Corresponding to the above method embodiment, a readable storage medium is further provided in this embodiment. A readable storage medium described below and a method for acquiring engineering control points along a line described above can be referred to each other correspondingly.
[0167] A readable storage medium has a computer program stored thereon. When the computer program is executed by a processor, the steps of the method for acquiring engineering control points along a line in the above method embodiment are implemented.
[0168] The readable storage medium may specifically be various readable storage media such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc that can store program codes.
[0169] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention may have various changes and modifications. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.
[0170] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims
1. A method for obtaining engineering control points along a line, characterized in that, Including: Obtain a visual state space model; Based on the visual state space model, perform image extraction on preset multi-dimensional features to obtain image features of parallel branches; Perform image processing on the image features of the parallel branches to obtain a feature map of the parallel branches; Based on a preset decoder, perform feature extraction and fusion on the feature map of the parallel branches to obtain multi-scale features of the image; According to a preset network deep supervision method, perform feature extraction on the image features of the parallel branches and the multi-scale features of the image to obtain a ground object model along the line; Based on a preset loss function and a preset network deep supervision method, perform optimization and solution on the ground object model along the line to obtain an optimal ground object model along the line; Wherein the specific process of performing optimization and solution on the ground object model along the line by the preset loss function and the preset network deep supervision method includes: Obtain label contour information of a target image, and the label contour information of the target image includes true label contour values, a first internal pixel quantity, and a second internal pixel quantity; Based on the preset loss function, calculate the true label contour values to obtain a loss value; Based on the preset network deep supervision method, calculate the first internal pixel quantity and the second internal pixel quantity to obtain a sample result; Construct according to the loss value, the sample result, a preset weight parameter, and a regularization term of the preset weight parameter to obtain a ground object loss calculation formula; Based on the ground object loss calculation formula, perform optimization and solution on the ground object model along the line to obtain an optimal ground object model along the line; According to a preset overlay analysis method, solve the optimal ground object model along the line to obtain control points of the project along the line.
2. The method for obtaining control points of a project along the line according to claim 1, characterized in that the visual state space model includes a feature selection module, and performing image extraction on preset multi-dimensional features based on the visual state space model to obtain image features of parallel branches includes: Obtain images along the project line; Perform geometric correction and radiometric correction on the images along the project line to obtain a target image; Based on a preset ground object extraction network model, perform convolution and pooling operations on the target image to obtain initial features of the image; Perform multi-dimensional feature extraction on the initial features of the image to obtain multi-dimensional features; Based on the feature selection module, perform image extraction on the multi-dimensional features to obtain image features of parallel branches.
3. The method for obtaining control points of a project along the line according to claim 2, characterized in that the ground object extraction network model includes a deep residual network model, and performing multi-dimensional feature extraction on the initial features of the image to obtain multi-dimensional features includes: According to the deep residual network model, solve the initial features of the image to obtain an initial feature map; Perform deformable convolution processing on the initial feature map to obtain first branch features; Based on the feature selection module, process the first branch features to obtain processed first branch features; Perform deformable convolution processing on the processed first branch features to obtain second branch features; Perform integrated depth feature processing on the second branch features to obtain integrated second branch features; Perform deformable convolution processing on the first branch feature and the integrated second branch feature to obtain a third branch feature; Use the first branch feature, the second branch feature, and the third branch feature as multi-dimensional features.
4. The method for obtaining engineering control points along a line according to claim 2, characterized in that the decoder includes an attention module, and based on a preset decoder, feature extraction and fusion are performed on the feature maps of the parallel branches to obtain multi-scale features of the image, including: According to the preset separation ratio in the attention module, feature extraction and fusion are performed on the feature maps of the parallel branches to obtain the spatial features of the feature maps and the channel features of the feature maps; Perform convolution adjustment on the preset number of channels of the feature maps of the parallel branches to obtain the convolution channel number; Solve the feature maps of the parallel branches according to a preset logical function to obtain the attention features of the feature maps; Based on the attention module, calculate different convolution kernels and different dilation rates for the feature maps of the parallel branches to obtain a convolution data set; Based on the spatial features of the feature maps, the channel features of the feature maps, the convolution channel number, the attention features of the feature maps, and the convolution data set, construct multi-scale features of the image.
5. The method for obtaining engineering control points along a line according to claim 1, characterized in that, according to a preset overlay analysis method, solve the optimal along-line ground object model to obtain engineering control points along the line, including: Perform predictive remote sensing on the optimal along-line ground object model to obtain a first ground object category and a second ground object category; Construct according to the first ground object category and the second ground object category to obtain a ground object classification result; Based on a preset raster space data conversion library and a preset image processing library, solve the ground object classification result to obtain a ground object boundary; Based on a preset overlay analysis method, solve the ground object boundary and preset railway data to obtain engineering control points along the line.
6. An acquisition device for engineering control points along a line, characterized in that, Including: A first acquisition module for acquiring a visual state space model; A first processing module for performing image extraction on preset multi-dimensional features based on the visual state space model to obtain image features of parallel branches; A second processing module for performing image processing on the image features of the parallel branches to obtain feature maps of the parallel branches; A third processing module for performing feature extraction and fusion on the feature maps of the parallel branches based on a preset decoder to obtain multi-scale features of the image; A fourth processing module for performing feature extraction on the image features of the parallel branches and the multi-scale features of the image according to a preset network deep supervision method to obtain an along-line ground object model; A fifth processing module for optimizing and solving the along-line ground object model based on a preset loss function and a preset network deep supervision method to obtain an optimal along-line ground object model; Wherein the specific process of optimizing and solving the along-line ground object model by the preset loss function and the preset network deep supervision method includes: Obtain the label contour information of the target image, and the label contour information of the target image includes the true label contour value, the first internal pixel number, and the second internal pixel number; Calculate the true label contour value based on a preset loss function to obtain a loss value; Calculate the first internal pixel quantity and the second internal pixel quantity based on a preset network deep supervision method to obtain a sample result; Construct a ground object loss calculation formula according to the loss value, the sample result, a preset weight parameter, and a regularization term of the preset weight parameter; Optimize and solve the along-line ground object model based on the ground object loss calculation formula to obtain an optimal along-line ground object model; A sixth processing module, configured to solve the optimal along-line ground object model according to a preset overlay analysis method to obtain control points of the along-line project.
7. The device for obtaining control points of an along-line project according to claim 6, characterized in that the visual state space model includes a feature selection module, and the first processing module includes: A first acquisition unit, configured to acquire an image along the project; A first correction unit, configured to perform geometric correction and radiometric correction on the image along the project to obtain a target image; A first processing unit, configured to perform convolution and pooling operations on the target image based on a preset ground object extraction network model to obtain initial features of the image; A second processing unit, configured to perform multi-dimensional feature extraction on the initial features of the image to obtain multi-dimensional features; A third processing unit, configured to perform image extraction on the multi-dimensional features based on the feature selection module to obtain image features of a parallel branch.
8. The device for obtaining control points of an along-line project according to claim 6, characterized in that the sixth processing module includes: A first construction unit, configured to perform prediction remote sensing image on the optimal along-line ground object model to obtain a first ground object category and a second ground object category; A second construction unit, configured to construct according to the first ground object category and the second ground object category to obtain a ground object classification result; A third construction unit, configured to solve the ground object classification result based on a preset raster spatial data conversion library and a preset image processing library to obtain a ground object boundary; A fourth construction unit, configured to solve the ground object boundary and preset railway data based on a preset overlay analysis method to obtain control points of the along-line project.
9. An acquisition device for engineering control points along a line, characterized in that, including: A memory, configured to store a computer program; A processor, configured to implement the steps of the method for obtaining control points of an along-line project according to any one of claims 1 to 5 when executing the computer program.
10. A readable storage medium, characterized in that : A computer program is stored on the readable storage medium, and when the computer program is executed by a processor, the steps of the method for obtaining control points of an along-line project according to any one of claims 1 to 5 are implemented.
Citation Information
Patent Citations
Disordered point cloud curved surface reconstruction method based on adaptive learning neural network
CN115249298A