Surveying and mapping method and device based on deep learning

Abstract

The invention relates to the field of surveying and mapping, and relates to a surveying and mapping method and device based on deep learning, and the method comprises the steps: obtaining a surveying and mapping image, and carrying out the preprocessing of the surveying and mapping image, and obtaining a preprocessed image; inputting the preprocessed image into a local feature extraction layer in a surveying and mapping network based on deep learning to obtain local features; performing feature processing on the preprocessed image and the local features by using a global feature extraction layer in a deep learning-based surveying and mapping network to obtain global features; performing feature processing on the local features and the global features by adopting a feature increment processing layer in a deep learning-based surveying and mapping network to obtain feature differences; performing feature fusion on the local features, the global features and the feature differences by adopting a feature fusion layer in the surveying and mapping network to obtain fusion features; and outputting the fusion features by adopting an output layer in a deep learning-based surveying and mapping network to obtain a surveying and mapping result. The data processing efficiency and accuracy are greatly improved, manual intervention is reduced, and errors are reduced.

Description

Technical Field

[0001] The present disclosure relates to the technical field of civil engineering surveying and mapping, and in particular to a surveying and mapping method and device based on deep learning. Background Art

[0002] In the field of civil engineering, surveying and mapping technology is a basic and key link, which involves the precise measurement and recording of terrain, buildings and infrastructure. With the rapid development of remote sensing technology, geographic information system (GIS) and artificial intelligence, civil surveying and mapping is gradually shifting from traditional manual drawing and rule-based automated processing to data-driven intelligent analysis and processing. Traditional civil surveying and mapping methods rely on manual measurement and post-image processing, which are time-consuming and labor-intensive, and are easily affected by the operator's technical level and environmental conditions.

[0003] In view of this, this application is filed. Summary of the invention

[0004] In order to solve the above technical problems or at least partially solve the above technical problems, the embodiments of the present disclosure provide a surveying and mapping method and device based on deep learning, which can process a large amount of data and automatically complete data cleaning, classification and analysis, greatly improving the efficiency and accuracy of data processing, reducing human intervention and reducing errors.

[0005] In a first aspect, an embodiment of the present disclosure provides a surveying and mapping method based on deep learning, the method comprising:

[0006] Acquire a surveying and mapping image, and preprocess the surveying and mapping image to obtain a preprocessed image;

[0007] Inputting the preprocessed image into a local feature extraction layer in a deep learning-based mapping network to obtain local features;

[0008] Using a global feature extraction layer in a deep learning-based surveying and mapping network to perform feature processing on the preprocessed image and the local features to obtain global features;

[0009] Using a feature increment processing layer in a deep learning-based surveying and mapping network to perform feature processing on the local features and the global features to obtain feature differences;

[0010] Using a feature fusion layer in a deep learning-based surveying and mapping network to fuse the local features, the global features, and the feature differences to obtain fused features;

[0011] The output layer in the deep learning-based surveying and mapping network is used to output the fusion features to obtain the surveying and mapping results.

[0012] In a second aspect, the embodiments of the present disclosure further provide a surveying and mapping device based on deep learning, the device comprising:

[0013] A first acquisition module is used to acquire a surveying and mapping image and preprocess the surveying and mapping image to obtain a preprocessed image;

[0014] A second acquisition module is used to input the preprocessed image into a local feature extraction layer in a surveying and mapping network based on deep learning to obtain local features;

[0015] A third acquisition module is used to perform feature processing on the preprocessed image and the local features using a global feature extraction layer in a surveying and mapping network based on deep learning to obtain global features;

[0016] A fourth acquisition module is used to perform feature processing on the local features and the global features using a feature increment processing layer in a surveying and mapping network based on deep learning to obtain feature differences;

[0017] A fifth acquisition module is used to use a feature fusion layer in a surveying and mapping network based on deep learning to perform feature fusion on the local features, the global features and the feature differences to obtain fused features;

[0018] The sixth acquisition module is used to output the fusion features using the output layer in the surveying and mapping network based on deep learning to obtain the surveying and mapping results.

[0019] In a third aspect, an embodiment of the present disclosure further provides an electronic device, comprising: one or more processors; a storage device for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the deep learning-based surveying and mapping method as described above.

[0020] In a fourth aspect, an embodiment of the present disclosure further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the deep learning-based surveying and mapping method as described above.

[0021] The mapping method based on deep learning provided by the embodiment of the present disclosure can enhance the ability to analyze the features of the ground objects and improve the application effect of the mapping data by preprocessing the mapping image using the first formula to obtain the preprocessed image; then based on the specific combination and configuration of the local feature extraction layer, the global feature extraction layer and the feature fusion layer in the mapping network, an effective mechanism for fusing local and global features is established to enhance the ability of the mapping network to analyze complex scenes and significantly reduce the manual operation and time consumption required in the traditional mapping method. And by using advanced technologies such as dilated convolution and transposed convolution, it can effectively capture the subtle changes and edge details of the terrain, buildings and roads, thereby improving the accuracy of the mapping results. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the accompanying drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and the originals and elements are not necessarily drawn to scale.

[0023] Figure 1 This is a flowchart of a deep learning-based surveying and mapping method in an embodiment of the present disclosure.

[0024] Figure 2 Schematic diagram of the structure of a deep learning-based surveying and mapping device in an embodiment of the present disclosure;

[0025] Figure 3 It is a structural schematic diagram of an electronic device in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0026] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments described herein, which are instead provided for a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0027] It should be noted that the concepts such as "first" and "second" mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0028] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0029] Traditional surveying and mapping technologies, such as total station measurement and GPS positioning, have historically provided important data support for engineering construction. However, with the rapid development of science and technology, these technologies have gradually exposed some limitations. Insufficient accuracy, low efficiency and limited data processing capabilities are the main challenges facing traditional surveying and mapping technologies. First of all, in terms of accuracy, traditional surveying and mapping technologies usually rely on manual measurement tools for on-site data collection. Due to the limitations of human operating capabilities and environmental factors, it is difficult to ensure the absolute accuracy of the collected data. In complex civil engineering projects, especially urban construction, road construction and bridge engineering, high-precision terrain data and building information are required to provide a basis for design and construction. The shortcomings of traditional methods may lead to design errors or construction deviations, resulting in waste of resources and safety hazards. Secondly, efficiency issues are also a major shortcoming of traditional surveying and mapping methods. The traditional surveying and mapping process often involves multiple steps, including on-site surveys, data recording, drawing drawings, etc., and each link requires a lot of manpower and time. Especially in large-scale projects, the complex on-site environment and the changeable weather often lead to an extension of the surveying and mapping cycle, affecting the overall progress of the project. Therefore, improving the level of automation and intelligence in the surveying and mapping process and reducing human intervention have become urgent issues to be addressed in the industry.

[0030] Secondly, traditional data processing software is often unable to efficiently process these large-scale data, and information loss and errors are prone to occur during the processing process. Therefore, how to quickly and accurately analyze massive data and extract valuable information has become the key to improving the application level of civil surveying and mapping. In this context, the introduction of deep learning technology undoubtedly provides a new path for the processing and analysis of civil surveying and mapping data.

[0031] In response to the above problems, the embodiments of the present disclosure provide a surveying and mapping method based on deep learning, which can process large amounts of data and automatically complete data cleaning, classification and analysis, greatly improving the efficiency and accuracy of data processing, reducing human intervention and reducing errors.

[0032] Figure 1 : is a flow chart of a deep learning-based surveying and mapping method in an embodiment of the present disclosure. The method can be executed by a deep learning-based surveying and mapping device, which can be implemented in software and / or hardware, and can be configured in an electronic device. Figure 1 As shown, the method may specifically include the following steps:

[0033] S110: Acquire a surveying and mapping image, and preprocess the surveying and mapping image to obtain a preprocessed image.

[0034] Among them, remote sensing technologies such as satellite images and aerial photography can be used to collect high-quality mapping image data. Since remote sensing data is obtained under different lighting, atmospheric conditions and terrain conditions, processing parameters need to be adjusted according to different conditions. Secondly, the sensor itself may have systematic errors, such as radiation correction and geometric correction, which also need to be corrected. Preprocessing through the adaptive first formula can enhance the ability to analyze the characteristics of the ground objects and improve the application effect of remote sensing data.

[0035] In one embodiment, preprocessing the surveying and mapping image to obtain a preprocessed image includes:

[0036] The mapping image is preprocessed using a first formula to obtain a preprocessed image, wherein the first formula is expressed as:

[0037]

[0038] Where X represents the preprocessed image, P represents the survey image, j1 represents the pixel in the survey image, j2 represents the neighborhood pixel of j1, Cdf represents the cumulative distribution function, and L P1 Indicates the gray level number of j1, L P2 represents the gray level number of j2, μ represents the mean value of the sum of pixels of j1 and j2, k1 represents the filtering factor, E1 represents the radiation intensity factor, a represents the reflectivity factor of the atmosphere, θ2 represents the radiation angle factor, and θ1 represents the radiation brightness factor.

[0039] S120, inputting the preprocessed image into a local feature extraction layer in a surveying and mapping network based on deep learning to obtain local features.

[0040] Specifically, the deep learning surveying and mapping network includes a local feature extraction layer, a global feature extraction layer, a feature increment processing layer, a feature fusion layer and an output layer. The preprocessed image is first input into the local feature extraction layer in the deep learning-based surveying and mapping network to extract local features and obtain local features. The following expression can be used to determine the local features:

[0041]

[0042] in, represents the local features of the lth layer, PooL represents the average pooling operation, ReLU represents the activation function, DilatedConv represents the dilated convolution operation, X represents the preprocessed image, and W l local represents the weight of the local feature extraction layer at layer l, b l local Resconv represents the bias of the lth local feature extraction layer, and Resconv represents the deep residual variational convolution operation.

[0043] The local feature extraction layer in this embodiment focuses on capturing local details of terrain, buildings, roads, etc., such as edges, textures, and shapes. Using dilated convolutions can capture local contexts with a larger receptive field without significantly increasing the number of parameters. The local feature extraction layer effectively identifies edges in the image through convolution operations, which are key clues to distinguish different objects (such as the outline of a building or the boundary of a road). By capturing texture information in local areas, the local feature extraction layer can distinguish different types of surfaces and materials, for example, rough rock surfaces from smooth asphalt pavement. The local feature extraction layer can recognize various shapes, which is very important for distinguishing man-made structures of circular, rectangular or other shapes. Compared with traditional convolutions, dilated convolutions increase the receptive field, which enables the network to capture a wider range of contextual information without increasing parameters. This is crucial for understanding more complex local patterns while maintaining computational efficiency. Dilated convolutions allow the network to dynamically adjust its receptive field size to accommodate features of different resolutions, which is very useful for processing terrain features and man-made structures of different sizes.

[0044] S130. Use a global feature extraction layer in a deep learning-based surveying and mapping network to obtain global features from the preprocessed image and the local features.

[0045] On the basis of the above embodiments, after obtaining local features, global features are extracted. Global features contain key information such as overall terrain trends and the location of large structures. This information is crucial for understanding the overall layout of the surveying and mapping area. It is necessary to capture the global context information of the entire surveying and mapping area, such as the overall terrain trends and the location of large structures. Specifically, the global feature extraction layer in the surveying and mapping network based on deep learning can be used to extract features from the preprocessed image and local features to obtain global features. The global features can be determined using the following expression:

[0046]

[0047] in, represents the global features of the lth layer, StatPool represents the maximum pooling operation, tanh represents the hyperbolic tangent function, represents the local features of the lth layer, W l global represents the weight of the global feature extraction layer at layer l, b l global represents the bias of the lth global feature extraction layer, DeConv represents the transposed convolution operation, and X represents the preprocessed image.

[0048] The global feature extraction layer in this embodiment can capture the global contextual information of the entire surveying area, identify large-scale geographical features and man-made structures such as mountains, rivers, urban layouts, etc., and through the fully connected layer and tanh activation function, combined with maximum pooling and transposed convolution, it can extract global features and gradually reduce the spatial size of the features while maintaining sensitivity to the global structure.

[0049] S140. Perform feature processing on the local features and the global features using a feature increment processing layer in a deep learning-based surveying and mapping network to obtain feature differences.

[0050] In one embodiment, the feature increment processing layer in the deep learning-based surveying and mapping network is used to perform feature processing on the local features and the global features to obtain feature differences, including:

[0051] Performing dimension reduction processing on the global feature based on the dimension reduction factor in the feature increment processing layer to obtain a dimension-reduced global feature;

[0052] Performing dimension-upgrading processing on the local features based on the dimension-upgrading factors and residual connections in the feature increment processing layer to obtain dimension-upgrading local features;

[0053] Based on the reduced-dimensional global features and the increased-dimensional local features, feature differences are determined.

[0054] Exemplarily, the dimension reduction factor in the feature increment processing layer is used to reduce the dimension of the global feature using the following expression to obtain the dimension-reduced global feature, and the expression is:

[0055]

[0056] in, represents the global feature of dimensionality reduction, γ represents the dimensionality reduction feature factor, represents the global features of the lth layer, W l global represents the weight of the global feature extraction layer at layer l, b l global Represents the bias of the lth global feature extraction layer, AdaptivePool represents the adaptive pooling operation, and tanh represents the hyperbolic tangent function.

[0057] The global feature dimensionality reduction process in this embodiment helps to reduce the complexity of the model, avoid overfitting, and improve computational efficiency. Its adaptive pooling and hyperbolic tangent activation function maintain the importance of features while reducing the dimensionality. The dimensionality reduction factor and the weighted square sum help to adjust the scale of the features.

[0058] At the same time, the dimension-raising factor and residual connection in the feature increment processing layer are used to increase the dimension of the local features using the following expression to obtain the dimension-raising local features, which are expressed as follows:

[0059]

[0060] in, represents the local feature of dimension increase, σ represents the feature factor of dimension increase, and Residual represents the residual connection operation. represents the local features of the lth layer, W l local represents the weight of the local feature extraction layer at layer l, b l local Represents the bias of the local feature extraction layer of layer l, ReLU represents the activation function, and DilatedConv represents the dilated convolution operation.

[0061] In this embodiment, the local feature dimensionality increase processing can increase the expressiveness of the feature and capture more detailed information. By introducing the dimensionality increase factor and residual connection, the dimensionality increase operation enhances the expressiveness of the feature and allows the network to learn more complex patterns. The dilated convolution and activation function further enrich the local features.

[0062] After obtaining the reduced-dimensional global features and the increased-dimensional local features, the feature differences can be determined using the following expression:

[0063]

[0064] in, Indicates feature differences, The Euclidean norm of the feature variance.

[0065] This embodiment dynamically improves the feature extraction capability by utilizing the feature difference between the dimensionality-reduced global features and the dimensionality-upgraded local features, thereby improving the accuracy of the deep learning-based surveying and mapping network.

[0066] S150, using a feature fusion layer in a deep learning-based surveying and mapping network to fuse the local features, the global features, and the feature differences to obtain fused features.

[0067] On the basis of the above embodiments, a feature fusion layer can be used to fuse local features and global features as well as feature differences between local features and global features, further enriching the feature extraction capability of the surveying and mapping network and providing a more comprehensive surveying and mapping data representation. The fusion feature can be expressed as follows:

[0068]

[0069] in, represents the fusion feature of the lth layer, α represents the weighting factor, Upsample represents the upsampling operation, Downsample represents the downsampling operation, and ε represents the feature adjustment factor.

[0070] This embodiment balances the detail information of the up-dimensional local features and the context information of the down-dimensional global features through weighting factors, upsampling and downsampling operations, and generates fused features. The fused features contain rich spatial information and improve the ability of the surveying and mapping network to analyze complex scenes.

[0071] S160. Use the output layer in the deep learning-based surveying and mapping network to output the fusion features to obtain surveying and mapping results.

[0072] Specifically, the output layer in the deep learning-based mapping network can use the following expression to output the fusion features and output the final mapping results. The mapping results can be civil mapping results, such as accurate topographic and geological data, where the topographic and geological data can include building and road information. Then, the building and road information is used for urban planning, including land use planning, transportation planning, and urban infrastructure layout, to provide accurate topographic and geological data for construction projects, and to help designers and engineers with building design and construction planning. The expression of the mapping result is:

[0073]

[0074] Among them, Y out represents the mapping result, FO represents the Sigmoid function, W out represents the output layer weight, b out represents the output layer bias, DeConv represents the transposed convolution operation, Represents the fused features of the lth layer.

[0075] It should be noted that the deep learning surveying and mapping network used in this application is a surveying and mapping network trained with surveying and mapping images, and the surveying and mapping network training process will not be described in detail here.

[0076] The deep learning-based surveying and mapping method provided in this embodiment, through the specific combination and configuration of the local feature extraction layer, the global feature extraction layer and the feature fusion layer in the surveying and mapping network, integrates the effective mechanism of local and global features, enhances the ability of the surveying and mapping network to analyze complex scenes, and significantly reduces the manual operation and time consumption required in the traditional surveying and mapping method. And by using advanced technologies such as dilated convolution and transposed convolution, it can effectively capture the slight changes and edge details of terrain, buildings and roads, thereby improving the accuracy of surveying and mapping results.

[0077] Figure 2 FIG. 1 is a schematic diagram of the structure of a deep learning-based surveying and mapping device in an embodiment of the present disclosure. Figure 2 As shown, the device includes: a first acquisition module 210, a second acquisition module 220, a third acquisition module 230, a fourth acquisition module 240, a fifth acquisition module 250, and a sixth acquisition module 260.

[0078] A first acquisition module 210 is used to acquire a surveying and mapping image and preprocess the surveying and mapping image to obtain a preprocessed image;

[0079] A second acquisition module 220 is used to input the preprocessed image into a local feature extraction layer in a deep learning-based mapping network to obtain local features;

[0080] A third acquisition module 230 is used to perform feature processing on the pre-processed image and the local features using a global feature extraction layer in a surveying and mapping network based on deep learning to obtain global features;

[0081] A fourth acquisition module 240 is used to perform feature processing on the local features and the global features using a feature increment processing layer in a surveying and mapping network based on deep learning to obtain feature differences;

[0082] A fifth acquisition module 250 is used to perform feature fusion on the local feature, the global feature and the feature difference by using a feature fusion layer in a surveying and mapping network based on deep learning to obtain a fused feature;

[0083] The sixth acquisition module 260 is used to output the fusion feature using the output layer in the surveying and mapping network based on deep learning to obtain the surveying and mapping result.

[0084] Optionally, the first acquisition module 210 is further configured to preprocess the surveying and mapping image using a first formula to obtain a preprocessed image, wherein the first formula is expressed as:

[0085]

[0086] Where X represents the preprocessed image, P represents the survey image, j1 represents the pixel in the survey image, j2 represents the neighborhood pixel of j1, Cdf represents the cumulative distribution function, and L P1 Indicates the gray level number of j1, L P2 represents the gray level number of j2, μ represents the mean value of the sum of pixels of j1 and j2, k1 represents the filtering factor, E1 represents the radiation intensity factor, a represents the reflectivity factor of the atmosphere, θ2 represents the radiation angle factor, and θ1 represents the radiation brightness factor.

[0087] Wherein, the expression of the local feature is:

[0088]

[0089] in, represents the local features of the lth layer, PooL represents the average pooling operation, ReLU represents the activation function, DilatedConv represents the dilated convolution operation, X represents the preprocessed image, and W l local represents the weight of the local feature extraction layer at layer l, b l local Resconv represents the bias of the lth local feature extraction layer, and Resconv represents the deep residual variational convolution operation.

[0090] The expression of the global feature is:

[0091]

[0092] in, represents the global features of the lth layer, StatPool represents the maximum pooling operation, tanh represents the hyperbolic tangent function, represents the local features of the lth layer, W l global represents the weight of the global feature extraction layer at layer l, b l global represents the bias of the lth global feature extraction layer, DeConv represents the transposed convolution operation, and X represents the preprocessed image.

[0093] Optionally, the fourth acquisition module is further used to perform dimensionality reduction processing on the global feature based on the dimensionality reduction factor in the feature increment processing layer to obtain a dimensionality-reduced global feature;

[0094] Performing dimension-upgrading processing on the local features based on the dimension-upgrading factors and residual connections in the feature increment processing layer to obtain dimension-upgrading local features;

[0095] Based on the reduced-dimensional global features and the increased-dimensional local features, feature differences are determined.

[0096] Among them, the expression of the dimension reduction global feature is:

[0097]

[0098] in, represents the global feature of dimensionality reduction, γ represents the dimensionality reduction feature factor, represents the global features of the lth layer, W l global represents the weight of the global feature extraction layer at layer l, b l global Represents the bias of the lth global feature extraction layer, AdaptivePool represents the adaptive pooling operation, and tanh represents the hyperbolic tangent function.

[0099] The expression of the dimension-raising local feature is:

[0100]

[0101] in, represents the local feature of dimension increase, σ represents the feature factor of dimension increase, and Residual represents the residual connection operation. represents the local features of the lth layer, W l local represents the weight of the local feature extraction layer at layer l, b l local Represents the bias of the local feature extraction layer of layer l, ReLU represents the activation function, and DilatedConv represents the dilated convolution operation.

[0102] The expression of the feature fusion is:

[0103]

[0104] in, represents the fusion feature of the lth layer, α represents the weighting factor, Upsample represents the upsampling operation, Downsample represents the downsampling operation, and ε represents the feature adjustment factor.

[0105] The deep learning-based surveying and mapping device provided in the embodiment of the present disclosure can execute the steps in the deep learning-based surveying and mapping method provided in the method embodiment of the present disclosure, and the execution steps and beneficial effects are no longer repeated here.

[0106] Figure 3 Schematic diagram of the structure of an electronic device in the embodiment of the present disclosure. Figure 3 , which shows a structural schematic diagram of an electronic device 500 suitable for implementing the embodiments of the present disclosure. Figure 3 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0107] like Figure 3As shown, the electronic device 500 may include a processing device 501, a ROM 502, a RAM 503, a bus 504, an input / output (I / O) interface 505, an input device 506, an output device 507, a storage device 508, and a communication device 509. The processing device (e.g., a central processing unit, a graphics processor, etc.) 501 can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 502 or the program loaded from the storage device 508 to the random access memory (RAM) 503 to implement the mapping method based on deep learning as described in the present disclosure. In the RAM 503, various programs and data required for the operation of the electronic device 500 are also stored. The processing device 501, the ROM 502, and the RAM 503 are connected to each other through the bus 504. The input / output (I / O) interface 505 is also connected to the bus 504.

[0108] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a non-transitory computer-readable medium, and the computer program contains a program code for executing the method shown in the flowchart, thereby implementing the surveying and mapping method based on deep learning as described above. In such an embodiment, the computer program can be downloaded and installed from the network through the communication device 509, or installed from the storage device 508, or installed from the ROM 502. When the computer program is executed by the processing device 501, the above-mentioned functions defined in the method of the embodiment of the present disclosure are executed.

[0109] It should be noted that the computer-readable medium disclosed above may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, device or device. In the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0110] The computer-readable medium may be included in the electronic device; or it may exist independently without being assembled into the electronic device. The computer-readable medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device: obtains a surveying image, and preprocesses the surveying image to obtain a preprocessed image; inputs the preprocessed image into a local feature extraction layer in a surveying and mapping network based on deep learning to obtain local features; uses a global feature extraction layer in a surveying and mapping network based on deep learning to perform feature processing on the preprocessed image and local features to obtain global features; uses a feature increment processing layer in a surveying and mapping network based on deep learning to perform feature processing on local features and global features to obtain feature differences; uses a feature fusion layer in a surveying and mapping network based on deep learning to perform feature fusion on local features, global features and feature differences to obtain fused features; uses an output layer in a surveying and mapping network based on deep learning to output fused features to obtain surveying and mapping results.

[0111] Optionally, when the above one or more programs are executed by the electronic device, the electronic device may also execute other steps described in the above embodiments.

[0112] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0113] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present disclosure is not limited to the technical solutions formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present disclosure (but not limited to) by each other to form a technical solution.

Claims

1. A surveying and mapping method based on deep learning, characterized in that: The method comprises: Acquire a surveying and mapping image, and preprocess the surveying and mapping image to obtain a preprocessed image; Inputting the preprocessed image into a local feature extraction layer in a deep learning-based mapping network to obtain local features; Using a global feature extraction layer in a deep learning-based surveying and mapping network to perform feature processing on the preprocessed image and the local features to obtain global features; The feature increment processing layer in the deep learning-based mapping network is used to process local features and global features to obtain feature differences; Using a feature fusion layer in a deep learning-based surveying and mapping network to fuse the local features, global features, and feature differences to obtain fused features; The output layer in the deep learning-based surveying and mapping network is used to output the fusion features to obtain the surveying and mapping results.

2. The method according to claim 1, characterized in that Preprocessing the surveying and mapping image to obtain a preprocessed image includes: The mapping image is preprocessed using a first formula to obtain a preprocessed image, wherein the first formula is expressed as: Where X represents the preprocessed image, P represents the survey image, j1 represents the pixel in the survey image, j2 represents the neighborhood pixel of j1, Cdf represents the cumulative distribution function, and L P1 Indicates the gray level number of j1, L P2 represents the gray level number of j2, μ represents the mean value of the sum of pixels of j1 and j2, k1 represents the filtering factor, E1 represents the radiation intensity factor, a represents the reflectivity factor of the atmosphere, θ2 represents the radiation angle factor, and θ1 represents the radiation brightness factor.

3. The method according to claim 1, characterized in that The expression of the local feature is: in, represents the local features of the lth layer, PooL represents the average pooling operation, ReLU represents the activation function, DilatedConv represents the dilated convolution operation, X represents the preprocessed image, and W l local represents the weight of the local feature extraction layer at layer l, b l local Resconv represents the bias of the lth local feature extraction layer, and Resconv represents the deep residual variational convolution operation.

4. The method according to claim 1, characterized in that: The expression of the global feature is: in, represents the global features of the lth layer, StatPool represents the maximum pooling operation, tanh represents the hyperbolic tangent function, represents the local features of the lth layer, W l global represents the weight of the global feature extraction layer at layer l, b l global represents the bias of the lth global feature extraction layer, DeConv represents the transposed convolution operation, and X represents the preprocessed image.

5. The method according to claim 1, characterized in that The feature increment processing layer in the deep learning-based mapping network is used to perform feature processing on the local features and the global features to obtain feature differences, including: Performing dimensionality reduction processing on the global features based on the dimensionality reduction factors in the feature increment processing layer to obtain dimensionality-reduced global features; Performing dimension-upgrading processing on the local features based on the dimension-upgrading factors and residual connections in the feature increment processing layer to obtain dimension-upgrading local features; Based on the dimension-reduced global features and the dimension-upgraded local features, feature differences are determined.

6. The method according to claim 5, characterized in that The expression of the dimension reduction global feature is: in, represents the global feature of dimensionality reduction, γ represents the dimensionality reduction feature factor, represents the global features of the lth layer, W l global represents the weight of the global feature extraction layer at layer l, b l global Represents the bias of the lth global feature extraction layer, AdaptivePool represents the adaptive pooling operation, and tanh represents the hyperbolic tangent function.

7. The method according to claim 6, characterized in that The expression of the dimension-raising local feature is: in, represents the local feature of dimension increase, σ represents the feature factor of dimension increase, and Residual represents the residual connection operation. represents the local features of the lth layer, W l local represents the weight of the local feature extraction layer at layer l, b l local Represents the bias of the local feature extraction layer of layer l, Re;U represents the activation function, and DilatedConv represents the dilated convolution operation.

8. The method according to claim 7, characterized in that The expression of the characteristic difference is: in, Indicates feature differences, The Euclidean norm of the feature differences.

9. The method according to claim 8, characterized in that The expression of the feature fusion is: in, represents the fusion feature of the lth layer, α represents the weighting factor, Upsample represents the upsampling operation, Downsamplw represents the downsampling operation, and ε represents the feature adjustment factor.

10. A surveying and mapping device based on deep learning, characterized in that: The device comprises: A first acquisition module is used to acquire a surveying and mapping image and preprocess the surveying and mapping image to obtain a preprocessed image; A second acquisition module is used to input the preprocessed image into a local feature extraction layer in a surveying and mapping network based on deep learning to obtain local features; A third acquisition module is used to perform feature processing on the preprocessed image and the local features using a global feature extraction layer in a surveying and mapping network based on deep learning to obtain global features; A fourth acquisition module is used to perform feature processing on the local features and the global features using a feature increment processing layer in a surveying and mapping network based on deep learning to obtain feature differences; A fifth acquisition module is used to use a feature fusion layer in a surveying and mapping network based on deep learning to perform feature fusion on the local features, the global features and the feature differences to obtain fused features; The sixth acquisition module is used to output the fusion features using the output layer in the surveying and mapping network based on deep learning to obtain the surveying and mapping results.

Images