Deep learning based 1:2000 topographic map feature recognition and extraction method

Abstract

The present application relates to a 1:2000 topographic map feature recognition and extraction method based on deep learning, and relates to the technical field of geographic information, comprising: obtaining a first feature symbol of a 1:2000 topographic map in response to a circled area input by a user end; obtaining a first feature type matching result; obtaining a second feature type matching result when the first feature type matching result is empty; performing size reduction when the maximum size of the feature symbol size identifier is greater than or equal to 0.09 cm to obtain the actual size of the feature; marking the actual size of the feature as empty when the maximum size of the feature symbol size identifier is less than 0.09 cm; and sending the second feature type matching result, the actual size of the feature and the feature symbol position identifier to the user end. The present application solves the technical problems of insufficient accuracy and adaptability of existing topographic map feature recognition and extraction, and achieves the technical effects of improving the accuracy and applicability of topographic map feature recognition and information extraction.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of geographic information, and particularly relates to a 1:2000 topographic map feature recognition and extraction method based on deep learning. BACKGROUND

[0002] With the continuous development of geographic information systems, topographic map feature unmanned recognition is widely used in the fields of traffic navigation, map culture display, etc. In the prior art, topographic map feature recognition usually adopts a method based on map annotation symbols, that is, an association database of annotation symbols and feature information is pre-set, and then the map annotation symbols are extracted, and the association database is combined to realize the recognition and extraction of feature type, distribution, size, etc.

[0003] However, the above prior art scheme still has the following problems: first, topographic maps are diverse, and topographic maps of different sources and different periods have great differences in representation, which leads to insufficient applicability of the recognition method based on a single association database. Second, even for annotation symbols representing the same feature meaning, there may be slight differences in size and shape, and such non-standardized symbols are difficult to accurately match and recognize by the association database. Third, limited by the establishment cost and updating efficiency of the association database, the existing scheme is difficult to cover all possible annotation symbols, and cannot realize recognition and information extraction when facing unknown symbols. The above problems lead to insufficient accuracy and adaptability of the existing scheme in topographic map feature recognition and extraction, and it is difficult to meet the actual application requirements of topographic map feature unmanned recognition. SUMMARY

[0004] The present application provides a 1:2000 topographic map feature recognition and extraction method based on deep learning to solve the technical problem that the accuracy and adaptability of topographic map feature recognition and extraction are insufficient due to the factors of diverse topographic maps, difficulty in recognizing non-standardized annotation symbols, and inability to recognize unknown symbols in the prior art topographic map feature recognition and extraction method.

[0005] The technical scheme for solving the above technical problem is as follows:

[0006] The application provides a 1:2000 topographic map feature recognition and extraction method based on deep learning, comprising: obtaining a first feature symbol of a 1:2000 topographic map in response to a circled area input by a user end, wherein the first feature symbol has a feature symbol size identifier and a feature symbol position identifier; processing the first feature symbol by a symbol description twin network in combination with a symbol description database to obtain a first feature type matching result; when the first feature type matching result is empty, performing adjacent symbol analysis on the first feature symbol in combination with a topographic map sharing platform to obtain a second feature type matching result; when the maximum size of the feature symbol size identifier is greater than or equal to 0.09 cm, performing size reduction according to a 1:2000 scale to obtain a feature actual size; when the maximum size of the feature symbol size identifier is less than 0.09 cm, marking the feature actual size as empty; and sending the second feature type matching result, the feature actual size and the feature symbol position identifier to the user end.

[0007] The application has the following beneficial effects:

[0008] In response to a circled area input by a user end, a first feature symbol of a 1:2000 topographic map is obtained, wherein the first feature symbol has a feature symbol size identifier and a feature symbol position identifier; the map annotation symbol circled by the user is obtained as a to-be-recognized object, and the size and position information thereof are extracted to provide input data for subsequent recognition and information extraction; the first feature symbol is processed by a symbol description twin network in combination with a symbol description database to obtain a first feature type matching result; the symbol description twin network is introduced, and a deep learning algorithm is used to learn and match the features of the annotation symbol, and the symbol description database is combined to realize preliminary identification of the feature type; when the first feature type matching result is empty, adjacent symbol analysis is performed on the first feature symbol in combination with a topographic map sharing platform to obtain a second feature type matching result; for the feature symbol that fails to be matched, the adjacent symbol analysis is used for re-matching to improve the recognition capability of non-standardized and unknown symbols; when the maximum size of the feature symbol size identifier is greater than or equal to 0.09 cm, size reduction is performed according to a 1:2000 scale to obtain a feature actual size; when the maximum size of the feature symbol size identifier is less than 0.09 cm, the feature actual size is marked as empty; the symbol size is reduced according to the map scale to obtain the actual size of the real feature, and the feature with too small size is filtered out to improve the accuracy of information extraction; and the second feature type matching result, the feature actual size and the feature symbol position identifier are sent to the user end, thereby improving the accuracy and applicability of topographic map feature recognition and extraction. BRIEF DESCRIPTION OF DRAWINGS

[0009] Figure 1A flowchart of a 1:2000 topographic map feature recognition and extraction method based on deep learning provided by the present application is shown in the figure.

[0010] Figure 2 A structural schematic diagram of an electronic device provided by the present application is shown in the figure.

[0011] Figure 3 A structural schematic diagram of a computer readable storage medium provided by the present application is shown in the figure.

[0012] In the drawings, the components represented by the respective reference numerals are as follows:

[0013] The electronic device 100, the memory 110, the processor 120, the first computer program 111, the computer readable storage medium 200, and the second computer program 211. DETAILED DESCRIPTION

[0014] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0015] In the description of the present application, the terms "first", "second" are used only for the purpose of description, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include one or more of the features. In the description of the present application, the meaning of "multiple" is two or more, unless otherwise specifically limited.

[0016] In the description of the present application, the term "for example" is used to indicate "as an example, illustration or explanation". Any embodiment described as "for example" in the present application is not necessarily interpreted as more preferred or more advantageous than other embodiments. The following description is given in order to enable any person skilled in the art to implement and use the present application. In the following description, details are listed for the purpose of explanation. It should be understood that those skilled in the art can realize the present application without using these specific details. In other examples, well-known structures and processes will not be described in detail in order to avoid unnecessary details making the description of the present application obscure. Therefore, the present application is not intended to be limited to the shown embodiments, but is consistent with the broadest scope of principles and features disclosed.

[0017] Embodiment one:

[0018] As Figure 1As shown, the embodiment of the present application provides a 1:2000 topographic map feature recognition and extraction method based on deep learning, comprising:

[0019] S100: obtaining a first feature symbol of the 1:2000 topographic map in response to the user's input of the circled area, wherein the first feature symbol has a feature symbol size identifier and a feature symbol position identifier.

[0020] Specifically, the user can open an image browsing interface that stores the 1:2000 topographic map to be identified through a user terminal device such as a smart phone, a tablet computer or a PC, etc., and then manually or through other human-computer interaction methods, circulate the area of interest on the image browsing interface to obtain the circled area. Subsequently, in response to the user's circled area, the image content in the circled area is taken as the target data for subsequent identification. Then, the feature symbols are obtained from the circled area in sequence, and each obtained feature symbol is taken as the first feature symbol. The so-called feature symbol refers to the symbolic expression of various types of feature entities (such as buildings, roads, water systems, etc.) in 1:2000 topographic map using specific graphics. Different types of features are usually represented by different symbols.

[0021] When extracting the first feature symbol, two attributes associated with it are also obtained, which are feature symbol size identifier and feature symbol position identifier. Among them, the feature symbol size identifier represents the size information of the symbol in the graphics, and the feature symbol position identifier represents the spatial position information of the symbol. These two attribute information will be used to restore the actual size of the feature and determine its spatial position in the subsequent steps.

[0022] By responding to the user's circled input, the first feature symbol representing the feature entity is obtained from the target area of the 1:2000 topographic map, and the size and position attribute information related to the symbol are extracted, which prepares for further identifying the feature type represented by the symbol and obtaining the real information of the feature.

[0023] S200: processing the first feature symbol through the symbol description twin network, combining the symbol description database to obtain the first feature type matching result.

[0024] Specifically, since different 1:2000 topographic maps may come from different draftsmen, even if the same feature type is represented, the shape and color of the feature symbol used may also have some differences. In order to improve the accuracy and robustness of feature symbol recognition, the first feature symbol is processed through the symbol description twin network.

[0025] The so-called twin network is a neural network structure containing two sub-networks with the same structure and parameters. For example, the symbol description twin network contains two parallel feature extraction channels and a post-feature comparison channel. The first ground feature symbol to be identified and the reference ground feature symbol randomly selected from the symbol description database are input into the two feature extraction channels respectively, the shape and color features of each are extracted, and then the feature comparison channel calculates the similarity of the two sets of features. If the shape and color similarity both exceed the preset threshold, the class of the reference ground feature symbol is taken as the recognition result, i.e. the first ground feature type matching result.

[0026] The symbol description database is a data set storing the shape, color and corresponding ground feature class annotation information of various standard ground feature symbols. By continuously searching the symbol examples in the symbol description database, the most similar one to the first ground feature symbol can be found, so as to know the ground feature type represented by the first ground feature symbol. If no matching item with both shape and color similarity meeting the requirements is found after traversing the symbol description database, the first ground feature type matching result is set to empty, indicating that the semantic class of the first ground feature symbol cannot be directly recognized at present.

[0027] By using the symbol description twin network, the reference symbol with the most similar shape and color is matched from the symbol description database, the ground feature type corresponding to the first ground feature symbol is obtained as the first ground feature type matching result, which can effectively overcome the problem of different expression of symbols in different topographic maps and improve the generalization ability of ground feature symbol recognition.

[0028] S300: When the first ground feature type matching result is empty, the first ground feature symbol is analyzed in combination with the topographic map sharing platform, and the second ground feature type matching result is obtained.

[0029] Specifically, if the first ground feature type matching result is empty after processing by the symbol description twin network, it means that only relying on the shape and color of the symbol cannot determine the ground feature class to which the ground feature symbol belongs. At this time, the first ground feature symbol is analyzed in combination with the topographic map sharing platform, and the second ground feature type matching result is obtained.

[0030] Since the 1:2000 topographic map usually covers the terrain information within a certain area, the ground feature types represented by the ground feature symbols in adjacent positions in the map usually have certain relevance. For example, there are usually bridges near rivers, and roads are usually connected around buildings. Accordingly, the context semantic information formed by the adjacent symbols can be used to indirectly infer the ground feature symbol that cannot be directly recognized.

[0031] To achieve the above object, a topographic map sharing platform is introduced. The topographic map sharing platform is a topographic map data sharing and service platform based on cloud technology, stores a large amount of 1:2000 topographic map data of different regions and different periods, and properly classifies and semantically annotates the data. Through learning and analysis of massive topographic map data, the topographic map sharing platform accumulates prior knowledge about the spatial correlation rules between different ground object types.

[0032] In specific implementation, the first ground object symbol is submitted to the topographic map sharing platform, and a plurality of adjacent symbols similar in shape and color to the first ground object symbol are searched in the topographic map sharing platform. The ground object type information of the adjacent symbols is summarized and counted, and the ground object type with the highest occurrence frequency is taken as the second ground object type matching result. For example, if the symbol A to be recognized is similar to symbols B, C and D in the topographic map sharing platform, and the most frequently appearing ground object in the neighborhood of the three symbols is a "road", it can be inferred that symbol A is also likely to represent a road.

[0033] By fully utilizing the topographic map sharing platform, the symbol semantic information around the first ground object symbol is analyzed and mined, and then the ground object type of the symbol itself is reasonably inferred, solving the problem that the symbol itself is difficult to accurately recognize, and improving the accuracy of ground object symbol recognition as a whole.

[0034] S400: When the maximum size of the ground object symbol size identifier is greater than or equal to 0.09 cm, the size is reduced according to the 1:2000 scale to obtain the actual size of the ground object.

[0035] Specifically, in the 1:2000 topographic map, each ground object symbol is usually drawn according to a certain scaling ratio so as to clearly express on the map. However, in the application link, more attention is often paid to the real size information of the ground object. Therefore, it is necessary to reduce the symbol size to the actual size according to the drawing scale of the topographic map.

[0036] In 1:2000 scale, 1 cm on the map represents 20 meters in the field, that is, the ratio of the size of the map to the actual size is 1:2000. Combined with practical experience, 0.09 cm is taken as the threshold value, that is, when the maximum size of the size mark of the ground symbol (referring to the maximum value of the length, width, height, etc. of the symbol) is greater than or equal to 0.09 cm, it is considered that the real size of the ground symbol is large enough to be worthy of reduction calculation; when the maximum size is less than 0.09 cm, it can be ignored and not reduced. In this way, a balance can be achieved between the integrity of the reduction information and the calculation efficiency. For example, if the width of the first ground symbol to be processed is 0.13 cm and the height is 0.09 cm, the maximum size (0.13 cm) exceeds the threshold value and needs to be reduced. Multiply 0.13 by 2000 to get the actual width of the ground as 260 cm, that is, 2.6 meters. The height size is equal to the threshold value and can also be reduced to 1.8 meters according to the above method.

[0037] By using the method of multiplying the reciprocal of the scale for ground symbols with a size greater than a certain threshold value, the actual size of the ground is calculated and reduced, which can accurately obtain the real geometric information of the ground symbol on one hand, and on the other hand, by setting the threshold value, the possible small errors under the limitation of the map accuracy are excluded, making the reduction result more reliable.

[0038] S500: When the maximum size of the size mark of the ground symbol is less than 0.09 cm, the actual size of the ground is marked as null.

[0039] Specifically, as described above, 0.09 cm is set as the threshold value for determining whether to reduce the size of the ground symbol. When the maximum size of the ground symbol is less than the threshold value, it is considered that the real ground corresponding to the symbol is either very small or difficult to distinguish from the adjacent ground, and therefore it is not meaningful to calculate and reduce the real size of the ground. Therefore, the actual size information of such ground symbol is directly marked as null and not reduced.

[0040] For example, if the width of the ground symbol to be processed is 0.08 cm and the height is 0.06 cm, the maximum size (0.08 cm) is less than the threshold value 0.09 cm, and therefore the actual width and height of the ground symbol are both marked as null. The null can be represented by a specific character such as "null" or a convention value such as "-1", as long as it can be distinguished from the valid size value.

[0041] By not reducing the actual size of the ground symbol with too small size, only marking it as null, the balance between information quality and processing efficiency can be achieved, which better supports subsequent applications.

[0042] S600: send the second feature type matching result, the feature actual size and the feature symbol position identifier to the user terminal.

[0043] Specifically, the second feature type matching result represents the real feature type represented by the first feature symbol, the feature actual size reflects the real geometric size of the feature, and the feature symbol position identifier indicates the spatial position. The three contents constitute a comprehensive and accurate semantic description of a feature symbol, which contains the feature type, size, position and other key information that the user is interested in.

[0044] The above-mentioned three contents extracted by recognition are packaged and sent to the user terminal that initiates the recognition request in real time through network transmission. On the user terminal side, the received feature semantic information can be presented to the user through a specially designed information display interface. For example, in the topographic map browsing interface, the feature type text identified by the feature symbol is overlaid or labeled; the length, width, height and other size information of the feature is displayed in the form of pop-up window, dialog box, etc.; the center point position of the feature symbol is accurately marked in the spatial coordinate system of the map, etc. The user can intuitively and accurately master the semantic connotation of the original topographic map combined with the recognition result, greatly improving the readability and usability of the graphical information. Further, in some application scenarios, the user terminal can also convert the received feature semantic information into other common data formats, such as XML, JSON, etc., for sharing and pushing between different systems and platforms.

[0045] By sending the results of feature symbol recognition and size restoration to the user terminal through the network and displaying them, the technical effect of improving the accuracy and applicability of topographic map feature recognition and information extraction is achieved.

[0046] Further, the embodiments of the present application also include:

[0047] S110: segmenting the 1:2000 topographic map according to the circled area to obtain a segmented digital image;

[0048] S120: performing semantic segmentation on the segmented digital image through a multi-layer convolution symbol detection model to extract a feature symbol list, a feature symbol size identifier list and a feature symbol position identifier list;

[0049] S130: the first feature symbol belongs to the feature symbol list, the feature symbol size identifier belongs to the feature symbol size identifier list, and the feature symbol position identifier belongs to the feature symbol position identifier list.

[0050] In a feasible implementation, the specific implementation of S100 can be further refined as steps S110 to S130, so as to illustrate how to obtain the first feature symbol and its related attributes from the 1:2000 topographic map according to the circled area of the user.

[0051] First, the user's interested region circled on the terminal device by manual or other human-computer interaction is obtained, i.e., the circled region. Then, the circled region is taken as a cutting boundary, and an image segmentation operation is performed on the complete 1:2000 topographic map to obtain a sub-image, referred to as a segmented digital image. The segmented digital image only contains topographic map information within the circled region, reducing the data scale of subsequent recognition processing. Then, a multi-layer convolutional neural network model based on deep learning is used to detect and identify the ground feature symbols in the segmented digital image. The multi-layer convolutional neural network model takes the segmented digital image as input, and through a series of operations such as convolution, pooling, and activation, accurately segments the graphic area of each ground feature symbol from the complex image background. Further, the model can also calculate the geometric size parameters (such as length, width, and height) of each symbol area and the spatial position parameters in the image coordinate system. Finally, the model outputs three lists: a ground feature symbol list, a ground feature symbol size identifier list, and a ground feature symbol position identifier list, which respectively store the detected ground feature symbol graphics, corresponding size information, and position information, and prepare for subsequent processing.

[0052] Any ground feature symbol in the obtained ground feature symbol list can be used as a first ground feature symbol for subsequent processing. The first ground feature symbol can be specified by the user or selected randomly or according to certain rules by the system. At the same time, the ground feature symbol size identifier of the first ground feature symbol can be extracted from the ground feature symbol size identifier list, and the ground feature symbol position identifier of the first ground feature symbol can be extracted from the ground feature symbol position identifier list.

[0053] By first performing image segmentation on the original topographic map according to the user's circled region, then using a multi-layer convolutional symbol detection model designed for ground feature symbol recognition to identify the ground feature symbols in the segmented sub-image, and extracting the graphics, size, and position information of the ground feature symbols, and determining any symbol as a target symbol for subsequent processing, a basis is provided for subsequent ground feature recognition and extraction.

[0054] Further, the embodiments of the present application also include:

[0055] S121: setting a symbol detection size threshold of the 1:2000 topographic map, wherein the symbol detection size threshold is less than or equal to 0.33 mm;

[0056] S122: initializing a feature pyramid topology according to the symbol detection size threshold to obtain a plurality of initial feature pyramid topologies;

[0057] S123: training the plurality of initial feature pyramid topologies to obtain a plurality of ground feature symbol segmentation accuracies, wherein the ground feature symbol segmentation accuracy is represented by the similarity between the ground feature segmentation symbol and the ground feature preset supervision symbol;

[0058] S124: When the several ground object symbol segmentation accuracies are all less than or equal to the convergence accuracy threshold, performing feature pyramid topology optimization on the several initial feature pyramid topologies according to the several ground object symbol segmentation accuracies, to obtain a feature pyramid target topology greater than the convergence accuracy threshold;

[0059] S125: When the several ground object symbol segmentation accuracies have initial feature pyramid topologies greater than the convergence accuracy threshold, setting the initial feature pyramid topologies as the feature pyramid target topology;

[0060] S126: Generating the multi-layer convolutional symbol detection model according to the feature pyramid target topology.

[0061] In a preferred embodiment, detailed steps of constructing the multi-layer convolutional symbol detection model, i.e., steps S121 to S126, are provided to illustrate how to design and optimize the network structure and parameters of the detection model for 1:2000 topographic maps to achieve high-precision and robust symbol recognition.

[0062] First, a symbol detection size threshold of 1:2000 topographic maps is set, wherein the symbol detection size threshold is less than or equal to 0.33 mm. According to relevant standards, the minimum size requirement of ground object symbols in 1:2000 topographic maps is 0.33 mm. Based on this, the symbol detection size threshold specifically for this scale of topographic maps is set to provide prior knowledge for subsequent network training. Then, according to the symbol detection size threshold, the feature pyramid topology is initialized to obtain several initial feature pyramid topologies. The so-called feature pyramid is a multi-scale and hierarchical convolutional neural network structure that can extract image features at different receptive fields and resolutions. With the symbol detection size threshold as a constraint condition, a set of initial feature pyramid network topologies is randomly generated as a starting point for subsequent structure search. Subsequently, the several initial feature pyramid topologies are trained to obtain several ground object symbol segmentation accuracies, wherein the ground object symbol segmentation accuracy is represented using the similarity between the ground object segmentation symbols and the ground object preset supervision symbols. Specifically, the manually annotated topographic map dataset is used to train each initial topology structure, and the symbol segmentation accuracy is evaluated on the validation set. The definition of segmentation accuracy is based on the similarity measure between the segmented symbols and the manually annotated supervision symbols.

[0063] When the segmentation accuracies of the several ground object symbols are all less than or equal to the convergence accuracy threshold, the initial feature pyramid topologies are optimized according to the segmentation accuracies of the several ground object symbols, to obtain a feature pyramid target topology greater than the convergence accuracy threshold. Specifically, if the segmentation accuracies of the existing topological structures are all not ideal, the network layers, filter sizes and numbers and other hyperparameters of the topologies are iteratively optimized until the accuracy of a certain structure exceeds the preset convergence threshold, and the structure is determined as the feature pyramid target topology to adapt to the characteristics of the current data and tasks. When the segmentation accuracies of the several ground object symbols have an initial feature pyramid topology greater than the convergence accuracy threshold, the initial feature pyramid topology is set as the feature pyramid target topology. Specifically, if the segmentation accuracy of a certain initial topology has reached the convergence threshold, the initial topology does not need to be optimized and is directly determined as the feature pyramid target topology to save the calculation cost. Then, the corresponding multi-layer convolutional neural network model is generated using the obtained feature pyramid target topology, and the construction of the detection model is completed.

[0064] By initializing, training, evaluating and dynamically optimizing the feature pyramid topologies under the constraint of the specially designed size threshold, the optimal network structure is automatically searched, and then a customized detection model is generated, which maximizes the automation of the model design process while ensuring the detection accuracy and reducing the manual trial and error cost.

[0065] Further, the embodiments of the present application also include:

[0066] S1241: constructing a feature pyramid topology distance evaluation function, wherein the feature pyramid topology distance evaluation function is used to evaluate the distance characteristic value of any two feature pyramid topologies;

[0067] S1242: performing cluster analysis on the several initial feature pyramid topologies based on a topology distance threshold according to the feature pyramid topology distance evaluation function, to obtain multiple cluster initial feature pyramid topologies;

[0068] S1243: extracting the highest accuracy initial feature pyramid topology of the multiple cluster initial feature pyramid topologies respectively according to the segmentation accuracies of the several ground object symbols, to obtain multiple highest accuracy initial feature pyramid topologies;

[0069] S1244: setting the multiple highest accuracy initial feature pyramid topologies as multiple guide feature pyramid topologies, and performing guide mutation on the initial feature pyramid topologies not in the cluster to obtain updated feature pyramid topologies, wherein the guide mutation refers to shortening the distance between the guided feature pyramid topology and the guide feature pyramid topology;

[0070] S1245: training the updated feature pyramid topologies to obtain updated ground object symbol segmentation accuracies;

[0071] S1246: When the updated feature pyramid topology is greater than the convergence accuracy threshold, the updated feature pyramid topology is set as the feature pyramid target topology, otherwise the feature pyramid topology optimization cycle is executed.

[0072] In a preferred embodiment, the specific implementation of optimizing several initial feature pyramid topologies can be further refined as steps S1241 to S1246, through the introduction of feature pyramid topology distance evaluation function, cluster analysis and guided mutation, etc. mechanisms, to search the optimal network structure more intelligently and efficiently.

[0073] Firstly, a feature pyramid topology distance evaluation function is constructed to quantitatively depict the distance characteristic value between any two topology structures, representing the difference between the two topology structures. The feature pyramid topology distance evaluation function considers factors such as the number of layers of the topology structure, the number of filter types of each layer, and the size of the receptive field, etc. The comprehensive difference value is calculated by weighted summation, providing a distance measurement standard for subsequent cluster analysis. Then, using the constructed feature pyramid topology distance evaluation function, a topology distance threshold is used as the radius to cluster several initial feature pyramid topologies, and topologies with similar structures are divided into the same cluster to obtain multi-cluster initial feature pyramid topologies. Multi-cluster initial feature pyramid topologies are coarse-grained grouping of initial topologies, and topologies within the cluster share local optimal substructures, thereby helping to narrow the scope of subsequent optimization search and improving optimization efficiency. Subsequently, according to the feature symbol segmentation accuracy of several initial feature pyramid topologies, the highest segmentation accuracy topology is selected from each cluster as the highest accuracy initial feature pyramid topology in the corresponding cluster, obtaining multiple highest accuracy initial feature pyramid topologies, representing high-quality topology structures within each cluster, which can guide the optimization direction of other topologies.

[0074] Then, the multiple highest accuracy initial feature pyramid topologies are used as multiple guide feature pyramid topologies, and the structure parameters of the non-cluster initial feature pyramid topologies are adjusted with reference to the multiple guide feature pyramid topologies, so that they gradually approach the guide feature pyramid topologies under the definition of the feature pyramid topology distance evaluation function, completing the iterative optimization of the structure. The step size of guided mutation can be dynamically controlled to balance the optimization speed and exploration breadth. After guided mutation, a group of updated feature pyramid topologies are generated. Then, the feature symbol segmentation accuracy of the updated feature pyramid topologies is re-evaluated. Once the accuracy of a certain updated topology exceeds the preset convergence accuracy threshold, it is determined as the final feature pyramid target topology, and the optimization process ends; otherwise, the steps of clustering, guided mutation, etc. will continue to be iteratively executed until the convergence condition is met or the maximum number of iterations is reached.

[0075] By fusing distance measurement, clustering analysis and guided variation mechanism, an effective feature pyramid topology optimization strategy is formed, which can quickly lock the optimal network structure in the search space while taking into account the optimization of local details, and provides a strong guarantee for high-precision and adaptive construction of multi-layer convolutional symbol detection model.

[0076] Further, the feature pyramid topology distance evaluation function is:

[0077] X = [(x 11 ,x 12 ),…,(x i1 ,x i2 ),…,(x N1 ,x N2 )]

[0078] Y = [(y 11 ,y 12 ),…,(y i1 ,y i2 ),…,(y M1 ,y M2 )]

[0079]

[0080] Wherein, D(X,Y) represents the distance between two feature pyramid topologies, X represents the first feature pyramid topology, Y represents the second feature pyramid topology, x 11 , x 12 represent the dimension of the first layer convolution kernel of the first feature pyramid topology in two directions, x i1 , x i2 represent the dimension of the i-th layer convolution kernel of the first feature pyramid topology in two directions, x N1 , x N2 represent the dimension of the N-th layer convolution kernel of the first feature pyramid topology in two directions, N represents the total number of layers of the first feature pyramid topology, y 11 , y 12 represent the dimension of the first layer convolution kernel of the second feature pyramid topology in two directions, y i1 , y i2 represent the dimension of the i-th layer convolution kernel of the second feature pyramid topology in two directions, y M1 , y M2 represent the dimension of the M-th layer convolution kernel of the second feature pyramid topology in two directions, and M represents the total number of layers of the second feature pyramid topology.

[0081] In a preferred embodiment, the feature pyramid topology distance evaluation function constructed can adopt the following form:

[0082] X = [(x 11 ,x12 ),…,(x i1 ,x i2 ),…,(x N1 ,x N2 ]

[0083] Y=[(y 11 ,y 12 ),…,(y i1 ,y i2 ),…,(y M1 ,y M2 ]

[0084]

[0085] Firstly, define two feature pyramid topologies X and Y to be compared, where X consists of N layers of convolution kernels, and Y consists of M layers of convolution kernels. The two dimensions of the i-th layer of convolution kernels in X are denoted by x i1 and x i2 , and the two dimensions of the i-th layer of convolution kernels in Y are denoted by y i1 and y i2 . Then, the distance D(X, Y) between the feature pyramid topologies X and Y is defined as:

[0086]

[0087] where f(M, N) is a binary function, i.e.:

[0088]

[0089] where M-N is taken when M is not equal to N, and 1 is taken when M is equal to N. min(M, N) represents the smaller value of M and N.

[0090] The core idea of the above feature pyramid topology distance evaluation function is: first, use f(M, N) to measure the difference in the number of layers between the two feature pyramid topologies, the greater the difference in the number of layers, the farther the distance between the two topologies; second, for two topologies with the same number of layers, calculate the Euclidean distance of the convolution kernel dimension difference layer by layer, and accumulate to get the total distance measure. Therefore, the difference in the number of layers determines the order of magnitude of the topology distance, and the difference in the convolution kernel dimension determines the fine ordering between topologies with the same number of layers.

[0091] The above feature pyramid topology distance evaluation function is only one possible implementation, and its specific form can be adjusted according to actual needs. However, regardless of the form, the basic starting point is the same, that is, to consider the number of layers and the convolution kernel dimension of the feature pyramid topology, and to quantitatively characterize the difference between the feature pyramid topologies in the structural feature space, to provide an operational distance measure for clustering analysis and guided variation.

[0092] By providing a feature pyramid topological distance evaluation function, the hierarchical structure information of the feature pyramid topology is fully utilized, and the topological comparison is converted into a comprehensive measurement of the number of layers and convolution kernel dimensions, laying the foundation for building a multi-layer convolutional symbol detection model.

[0093] Furthermore, the symbol indicates that the twin network includes a first feature extraction channel, a second feature extraction channel, and a feature comparison channel, wherein the first feature extraction channel and the second feature extraction channel have the same structural topology and parameters. Obtaining the first feature type matching result includes:

[0094] S210: Inputting the first land feature symbol into the first feature extraction channel to obtain a shape feature and a color feature of the first land feature symbol;

[0095] S220: Inputting the first reference ground feature symbol randomly extracted from the symbol description database into the second feature extraction channel to obtain a second ground feature symbol shape feature and a second ground feature symbol color feature;

[0096] S230: Inputting the first land feature symbol shape feature, the first land feature symbol color feature, the second land feature symbol shape feature, and the second land feature symbol color feature into the feature comparison channel to obtain shape similarity and color similarity;

[0097] S240: When the shape similarity is greater than or equal to a shape similarity threshold, and the color similarity is greater than or equal to a color similarity threshold, setting the associated feature type of the first reference feature symbol as the first feature type matching result; otherwise, updating the first reference feature symbol according to the symbol description database for cyclic matching;

[0098] S250: When the symbol description database does not have the first land feature type matching result, the first land feature type matching result is set to empty.

[0099] In a preferred embodiment, through steps S210 to S250, the internal structure and working principle of the twin network are explained in detail with symbols.

[0100] Firstly, the symbol description twin network is composed of three sub-networks, which are the first feature extraction channel, the second feature extraction channel and the feature comparison channel. Among them, the first feature extraction channel and the second feature extraction channel adopt the same structure topology and parameters. Then, the first ground object symbol to be identified is input into the first feature extraction channel, and after a series of convolution, pooling and other operations, the shape feature and color feature of the symbol are extracted as the first ground object symbol shape feature and the first ground object symbol color feature. Among them, the shape feature describes the geometric appearance of the symbol, and the color feature reflects the color attribute of the symbol. At the same time, a reference ground object symbol is selected from the symbol description database, which is called the first reference ground object symbol, and is input into the second feature extraction channel. Similarly, the second feature extraction channel outputs the shape feature and color feature of the reference symbol as the second ground object symbol shape feature and the second ground object symbol color feature.

[0101] Subsequently, the shape features and color features of the first ground object symbol and the first reference ground object symbol are summarized, that is, the first ground object symbol shape feature, the first ground object symbol color feature, the second ground object symbol shape feature and the second ground object symbol color feature, which are input into the feature comparison channel. The feature comparison channel uses similarity measurement to calculate the similarity degree of the first ground object symbol shape feature and the second ground object symbol shape feature in shape, and the similarity degree of the first ground object symbol color feature and the second ground object symbol color feature in shape, to obtain two indexes of shape similarity and color similarity. Then, threshold judgment is performed on the shape similarity and the color similarity. If the shape similarity and the color similarity both exceed the preset threshold, that is, the shape similarity is greater than or equal to the shape similarity threshold, and the color similarity is greater than or equal to the color similarity threshold, it is considered that the first ground object symbol and the first reference ground object symbol belong to the same category, and at this time the associated ground object type marked by the first reference ground object symbol in the symbol description database is taken as the first ground object type matching result. Otherwise, it is considered that the current first reference ground object symbol is not similar enough, and the first reference ground object symbol needs to be updated, and steps S220 to S240 are repeatedly executed until a similar matching symbol is found or the entire symbol description database is traversed. If there is no reference ground object symbol in the symbol description database that can match the first ground object symbol, that is, the effective first ground object type matching result cannot be obtained finally, it is set to null, indicating that there is no category in the symbol description database that matches the first ground object symbol.

[0102] By using the symbol description twin network to identify the type of ground object symbol, the diversity and complexity of the ground object symbol can be effectively dealt with, and the precision and efficiency of symbol recognition can be greatly improved.

[0103] Further, the application also includes:

[0104] S310: set a shape similarity adjacency threshold and a color similarity adjacency threshold;

[0105] S320: based on the shape similarity adjacency threshold and the color similarity adjacency threshold, matching a plurality of land feature types of a plurality of adjacent land feature symbols for the first land feature symbol from the topographic map sharing platform;

[0106] S330: extracting a highest frequency land feature type of the plurality of land feature types as the second land feature type matching result.

[0107] In a preferred embodiment, the steps S310 to S330 further illustrate how to use the topographic map sharing platform to further analyze the adjacent symbols of the first land feature symbol to obtain the second land feature type matching result when the first land feature type matching result is empty.

[0108] First, two threshold parameters are set, which are the shape similarity adjacency threshold and the color similarity adjacency threshold. These two thresholds are used to retrieve adjacent symbols similar to the first land feature symbol in the topographic map sharing platform, but their values are usually greater than the shape similarity threshold and the color similarity threshold. This is because, in the absence of direct matching, appropriately relaxing the similarity requirements of adjacent symbols helps to obtain more reference information. Subsequently, based on the above two adjacency thresholds, adjacent land feature symbols similar in shape and color to the first land feature symbol are retrieved in the topographic map sharing platform. The adjacent land feature symbols obtained by the query and their corresponding land feature types are provided as candidate matching items to the next step. Then, the frequency of each land feature type appearing in all adjacent land feature symbols is counted, and the land feature type with the highest frequency is determined as the second land feature type matching result. Here, the prior assumption is used that spatially adjacent and similar feature symbols have a consistent tendency in their land feature types. Frequency statistics can eliminate the interference of individual abnormal matches to some extent and improve the confidence of the matching result.

[0109] By making full use of the vast amount of land feature symbol data and the implicit spatial topology relationship contained in the topographic map sharing platform when direct matching fails, the type of the first land feature symbol is inferred. This inference mechanism based on adjacent analysis makes up for the shortcomings of single symbol shape and color features, and can mine matching clues in a larger range to improve the success rate of recognition.

[0110] Further, the embodiments of the present application also include:

[0111] S321: based on the shape similarity adjacency threshold and the color similarity adjacency threshold, matching a plurality of land feature types of a plurality of adjacent land feature symbols for the first land feature symbol from the topographic map sharing platform;

[0112] S322: When the data amount of the primary abutment ground object symbol is less than or equal to the data amount threshold, based on the shape similarity abutment threshold and the color similarity abutment threshold, a secondary abutment ground object symbol is matched from the topographic map sharing platform by traversing the primary abutment ground object symbol;

[0113] S323: The secondary abutment ground object symbol and the primary abutment ground object symbol are added to the several abutment ground object symbols.

[0114] In a preferred embodiment, by introducing primary abutment and secondary abutment, the matching range of abutment symbols is expanded in a hierarchical and progressive manner.

[0115] Firstly, based on the shape similarity abutment threshold and the color similarity abutment threshold, the ground object symbol satisfying the shape similarity abutment threshold and the color similarity abutment threshold with the first ground object symbol is retrieved in the topographic map sharing platform, which is called the primary abutment ground object symbol. Then, based on the primary abutment matching, the secondary abutment is introduced. When the number of the primary abutment ground object symbol is small, i.e., less than a certain preset data amount threshold, the primary abutment ground object symbol is taken as the basis to further retrieve the ground object symbol satisfying the shape similarity abutment threshold and the color similarity abutment threshold with the primary abutment ground object symbol in the topographic map sharing platform, to obtain the secondary abutment ground object symbol, so as to expand the ground object symbol when the primary abutment ground object symbol is small, to obtain more abundant information. Subsequently, the obtained primary abutment ground object symbol and the obtained secondary abutment ground object symbol are combined as the several abutment ground object symbols of the first ground object symbol, to prepare for subsequent frequency statistical analysis.

[0116] By adopting the hierarchical abutment symbol matching strategy from near to far and from direct to indirect, the robustness and adaptability in dealing with complex ground object symbols are enhanced.

[0117] The 1:2000 topographic map ground object recognition and extraction method based on deep learning provided by the embodiment of the application has at least the following technical effects:

[0118] In response to the circled area input by the user end, the first feature symbol of the 1:2000 topographic map is obtained, wherein the first feature symbol has a feature symbol size identifier and a feature symbol position identifier, providing a basis for subsequent recognition and information extraction. The first feature symbol is processed through the symbol description twin network in combination with the symbol description database to obtain a first feature type matching result, thereby improving the recognition ability of complex and non-standardized symbols. When the first feature type matching result is empty, the first feature symbol is analyzed for adjacent symbols in combination with the topographic map sharing platform to obtain a second feature type matching result, further improving the recall rate of recognition and making up for the shortcomings of single feature matching. When the maximum size of the feature symbol size identifier is greater than or equal to 0.09 cm, the size is restored according to the 1:2000 scale to obtain the actual size of the feature. When the maximum size of the feature symbol size identifier is less than 0.09 cm, the actual size of the feature is marked as empty, and valuable size information is extracted, thereby improving the accuracy of the information extraction result. The second feature type matching result, the actual size of the feature and the location identifier of the feature symbol are sent to the user end, thereby improving the accuracy and adaptability of topographic map feature recognition and information extraction.

[0119] Example 2:

[0120] See also Figure 2 , Figure 2 Schematic diagram of an embodiment of an electronic device provided by an embodiment of the present invention. Figure 2 As shown, an electronic device 100 provided by an embodiment of the present invention includes a memory 110, a processor 120, and a first computer program 111 stored in the memory 110 and executable on the processor 120. When the processor 120 executes the first computer program 111, a 1:2000 topographic map feature recognition and extraction method based on deep learning is implemented.

[0121] Example 3:

[0122] See also Figure 3 , Figure 3 Schematic diagram of an embodiment of a computer-readable storage medium provided in an embodiment of the present invention. Figure 3 As shown, this embodiment provides a computer-readable storage medium 200 on which a second computer program 211 is stored. When the second computer program 211 is executed by a processor, a method for identifying and extracting features in a 1:2000 topographic map based on deep learning is implemented.

[0123] It should be noted that, in the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0124] Those skilled in the art will appreciate that embodiments of the present application can be devised for a variety of applications. It is intended that the present application be limited only by the scope of the appended claims, and it is intended that various modifications and alterations made by those skilled in the art be considered as within the scope of the present application. The embodiments of the present application will be described with reference to the attached drawings identified below.

[0125] The present application is described in reference to the drawings using a flowchart and / or a block diagram of the method, apparatus (system) and computer program product according to embodiments of the application. It will be understood that each block of the flowchart and / or block diagram, and combinations of blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks.

[0126] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks.

[0127] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks.

[0128] Although preferred embodiments of the application have been described, those skilled in the art will recognize that additional modifications and variations are possible without departing from the spirit or scope of the application.

[0129] Obviously, numerous modifications and variations of the present application are possible in light of the above teachings. It is therefore to be understood that within the scope of the application, the application can be practiced otherwise than as specifically described.

Claims

1. A 1:2000 topographic map feature recognition and extraction method based on deep learning, characterized by: include: In response to the circled area input by the user terminal, a first feature symbol of a 1:2000 topographic map is obtained, wherein the first feature symbol has a feature symbol size identifier and a feature symbol position identifier; Processing the first ground feature symbol through the symbol description twin network in combination with the symbol description database to obtain a first ground feature type matching result; When the first feature type matching result is empty, performing adjacent symbol analysis on the first feature symbol in combination with the topographic map sharing platform to obtain a second feature type matching result; When the maximum size of the feature symbol size mark is greater than or equal to 0.09 cm, the size is restored according to the 1:2000 scale to obtain the actual size of the feature; When the maximum size of the feature symbol size is less than 0.09 cm, the actual size of the feature is marked as empty; Sending the second feature type matching result, the actual size of the feature, and the feature symbol location identifier to the user terminal; The method of obtaining the first land feature symbol of the 1:2000 topographic map in response to the circled area input by the user terminal includes: Segmenting the 1:2000 topographic map according to the demarcated area to obtain a segmented digital image; Performing semantic segmentation on the segmented digital image through a multi-layer convolutional symbol detection model to extract a list of ground object symbols, a list of ground object symbol size identifiers, and a list of ground object symbol position identifiers; The first feature symbol belongs to the feature symbol list, the feature symbol size identifier belongs to the feature symbol size identifier list, and the feature symbol position identifier belongs to the feature symbol position identifier list; The steps of constructing the multi-layer convolutional symbol detection model include: Setting a symbol detection size threshold for the 1:2000 topographic map, wherein the symbol detection size threshold is less than or equal to 0.33 mm; Initializing the feature pyramid topology according to the symbol detection size threshold to obtain a plurality of initial feature pyramid topologies; Training the plurality of initial feature pyramid topologies to obtain a plurality of ground object symbol segmentation accuracy rates, wherein the ground object symbol segmentation accuracy rates are characterized by similarity between the ground object segmentation symbols and the ground object preset supervision symbols; When the plurality of ground object symbol segmentation accuracy rates are all less than or equal to a convergence accuracy rate threshold, performing feature pyramid topology optimization on the plurality of initial feature pyramid topologies according to the plurality of ground object symbol segmentation accuracy rates to obtain a feature pyramid target topology greater than the convergence accuracy rate threshold; When the segmentation accuracy rates of the plurality of ground object symbols are greater than the convergence accuracy threshold, setting the initial feature pyramid topology greater than the convergence accuracy threshold as the feature pyramid target topology; The multi-layer convolutional symbol detection model is generated according to the feature pyramid target topology.

2. The method according to claim 1, wherein The method further comprises: performing feature pyramid topology optimization on the plurality of initial feature pyramid topologies according to the plurality of ground object symbol segmentation accuracy rates to obtain a feature pyramid target topology having a value greater than the convergence accuracy threshold value, including: Constructing a feature pyramid topology distance evaluation function, wherein the feature pyramid topology distance evaluation function is used to evaluate the distance characteristic value between any two feature pyramid topologies; performing cluster analysis on the plurality of initial feature pyramid topologies based on a topological distance threshold according to the feature pyramid topology distance evaluation function to obtain multiple clusters of initial feature pyramid topologies; Extracting the highest accuracy initial feature pyramid topologies from the multiple clusters of initial feature pyramid topologies according to the plurality of ground feature symbol segmentation accuracies, and obtaining a plurality of highest accuracy initial feature pyramid topologies; The multiple highest accuracy initial feature pyramid topologies are set as multiple guiding feature pyramid topologies, and the initial feature pyramid topologies of non-cluster are subjected to guided mutation to obtain updated feature pyramid topologies, wherein the guided mutation refers to shortening the distance between the guided feature pyramid topology and the guiding feature pyramid topology; Training the updated feature pyramid topology to obtain an updated ground feature symbol segmentation accuracy; When the updated ground feature symbol segmentation accuracy is greater than the convergence accuracy threshold, the updated feature pyramid topology is set as the feature pyramid target topology, otherwise a feature pyramid topology optimization loop is executed.

3. The method according to claim 2, wherein The feature pyramid topological distance evaluation function is: X=[(x 11 ,x 12 ),…,(x i1 ,x i2 ),…,(x N1 ,x N2 )] And=[(and 11 ,and 12 ),…,(and i1 ,and i2 ),…,(and M1 ,and M2 )] Among them, D(X,Y) represents the distance between two feature pyramid topologies, X represents the first feature pyramid topology, Y represents the second feature pyramid topology, x 11 , x 12 Characterize the dimensions of the first convolution kernel in both directions of the first feature pyramid topology, x i1 , x i2 Characterizes the dimensions of the convolution kernel of the first layer of the first feature pyramid topology in two directions, x N1 , x N2 Represents the dimensions of the Nth convolution kernel in two directions of the first feature pyramid topology, N represents the total number of layers of the first feature pyramid topology, y 11 ,y 12 Characterize the dimensions of the first layer of convolution kernel in two directions of the second feature pyramid topology, y i1 ,y i2 Characterizes the dimensions of the convolution kernel of the i-th layer in two directions of the second feature pyramid topology, y M1 ,y M2 Represents the dimensions of the Mth convolution kernel in two directions of the second feature pyramid topology, and M represents the total number of layers of the second feature pyramid topology.

4. The method according to claim 1, wherein The symbol description twin network includes a first feature extraction channel, a second feature extraction channel, and a feature comparison channel. The first feature extraction channel and the second feature extraction channel have the same structural topology and parameters. The symbol description twin network is used to process the first feature symbol in combination with the symbol description database to obtain a first feature type matching result, including: Inputting the first ground object symbol into the first feature extraction channel to obtain the first ground object symbol shape feature and the first ground object symbol color feature; Inputting a first reference ground feature symbol randomly extracted from the symbol description database into the second feature extraction channel to obtain a second ground feature symbol shape feature and a second ground feature symbol color feature; Inputting the first land feature symbol shape feature, the first land feature symbol color feature, the second land feature symbol shape feature, and the second land feature symbol color feature into the feature comparison channel to obtain shape similarity and color similarity; When the shape similarity is greater than or equal to a shape similarity threshold, and the color similarity is greater than or equal to a color similarity threshold, setting the associated feature type of the first reference feature symbol as the first feature type matching result; otherwise, updating the first reference feature symbol according to the symbol description database for cyclic matching; When the symbol indicates that there is no matching result for the first land feature type in the database, the matching result for the first land feature type is set to empty.

5. The method according to claim 1, wherein When the first feature type matching result is empty, performing adjacent symbol analysis on the first feature symbol in conjunction with the topographic map sharing platform to obtain a second feature type matching result includes: Set the shape similarity adjacency threshold and color similarity adjacency threshold; Based on the shape similarity adjacency threshold and the color similarity adjacency threshold, matching the first feature symbol with a plurality of feature types of adjacent feature symbols from the topographic map sharing platform; The highest frequency ground feature type among the plurality of ground feature types is extracted and set as the second ground feature type matching result.

6. The method according to claim 5, wherein Based on the shape similarity adjacency threshold and the color similarity adjacency threshold, matching the first feature symbol with a plurality of feature types of adjacent feature symbols from the topographic map sharing platform, including: Based on the shape similarity adjacency threshold and the color similarity adjacency threshold, matching the first feature symbol with a primary adjacent feature symbol from the topographic map sharing platform; When the data amount of the first-level adjacent feature symbol is less than or equal to the data amount threshold, based on the shape similarity adjacency threshold and the color similarity adjacency threshold, traversing the first-level adjacent feature symbols from the topographic map sharing platform to match the second-level adjacent feature symbols; The secondary adjacent feature symbol and the primary adjacent feature symbol are added to the plurality of adjacent feature symbols.

Images