Survey data processing method and system applied to image analysis

Abstract

The application provides a surveying and mapping data processing method and system applied to image analysis, and relates to the technical field of artificial intelligence.In the application, a surveying and mapping image to be analyzed is extracted; network optimization operation is performed to form an optimized image analysis network; the optimized image analysis network is used to analyze the analysis possibility parameters of each image pixel in at least two restored mapping images; according to the analysis possibility parameters of each image pixel in the at least two restored mapping images, a target restored mapping image is selected from the at least two restored mapping images; and the to-be-determined image type information in the target restored mapping image is marked as target image type information of the surveying and mapping image to be analyzed.Based on the above, the reliability of surveying and mapping image analysis can be improved to a certain extent.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, in particular to a surveying and mapping data processing method and system applied to image analysis. BACKGROUND

[0002] In the field of surveying and mapping, image acquisition is one of the important measures. For the acquired images, based on some requirements, the images need to be analyzed to determine the type of the images (such as the type of abnormal or deformed regions, the type of the corresponding regions, etc.). However, in the prior art, there is a problem of low reliability in the process of surveying and mapping image analysis. SUMMARY

[0003] Therefore, the purpose of the present application is to provide a surveying and mapping data processing method and system applied to image analysis to improve the reliability of surveying and mapping image analysis to a certain extent.

[0004] To achieve the above-mentioned purpose, the embodiments of the present application adopt the following technical solutions:

[0005] A surveying and mapping data processing method applied to image analysis, comprising:

[0006] extracting a to-be-analyzed surveying and mapping image, which is formed based on a target image surveying and mapping operation on a target region;

[0007] forming an optimized image analysis network by performing a network optimization operation;

[0008] using the optimized image analysis network to analyze the analysis possibility parameters of each image pixel in at least two restored mapping images, the restored mapping images including the to-be-analyzed surveying and mapping image and pending image type information;

[0009] selecting a target restored mapping image from the at least two restored mapping images according to the analysis possibility parameters of each image pixel in the at least two restored mapping images;

[0010] labeling the pending image type information in the target restored mapping image as the target image type information of the to-be-analyzed surveying and mapping image, the target image type information being used to reflect the image type of the to-be-analyzed surveying and mapping image.

[0011] In some preferred embodiments, in the surveying and mapping data processing method applied to image analysis, the step of using the optimized image analysis network to analyze the analysis possibility parameters of each image pixel in at least two restored mapping images comprises:

[0012] performing data fusion operation on the to-be-analyzed survey image and the to-be-determined image category information to form one of the at least two restored survey images; and loading the one of the at least two restored survey images into the optimized image analysis network to mine key information feature representation of the one of the at least two restored survey images, and analyze the analysis possibility parameters of each image pixel in the one of the at least two restored survey images according to the key information feature representation of the one of the at least two restored survey images; or

[0013] loading the to-be-analyzed survey image into the optimized image analysis network to mine key information feature representation of the to-be-analyzed survey image; performing aggregation operation on the key information feature representation of the to-be-analyzed survey image and the key information feature representation of the to-be-determined image category information to form key information feature representation of one of the at least two restored survey images by using the optimized image analysis network; and analyzing the analysis possibility parameters of each image pixel in the one of the at least two restored survey images according to the key information feature representation of the one of the at least two restored survey images by using the optimized image analysis network.

[0014] In some preferred embodiments, in the above-mentioned survey data processing method applied to image analysis, the step of performing network optimization operation to form an optimized image analysis network comprises:

[0015] extracting a plurality of primary exemplary survey images;

[0016] analyzing the analysis possibility parameters of each image pixel in any one of the plurality of primary exemplary survey images by using a primary candidate neural network;

[0017] performing network optimization operation on the primary candidate neural network according to the analysis possibility parameters of each image pixel in the plurality of primary exemplary survey images to form a primary optimized neural network, wherein the primary optimized neural network comprises a primary key information mining sub-network;

[0018] mining the image key information feature representation corresponding to the any one of the plurality of primary exemplary survey images by using the primary key information mining sub-network;

[0019] performing network optimization operation on the primary key information mining sub-network according to the image key information feature representation corresponding to each of the plurality of primary exemplary survey images to form a senior key information mining sub-network corresponding to the primary key information mining sub-network;

[0020] According to the high-level key information mining sub-network, a corresponding optimized image analysis network is formed.

[0021] In some preferred embodiments, in the above-mentioned survey data processing method applied to image analysis, the step of performing network optimization operation on the primary key information mining sub-network according to the image key information feature representation corresponding to each of the plurality of primary exemplary survey images, to form a high-level key information mining sub-network corresponding to the primary key information mining sub-network, comprises:

[0022] According to the image key information feature representation corresponding to each of the plurality of primary exemplary survey images, the network learning cost parameter corresponding to each primary exemplary survey image is analyzed;

[0023] According to the network learning cost parameter corresponding to each primary exemplary survey image, the network learning cost parameter corresponding to the primary key information mining sub-network is analyzed;

[0024] According to the network learning cost parameter corresponding to the primary key information mining sub-network, the network optimization operation is performed on the primary key information mining sub-network to form a high-level key information mining sub-network corresponding to the primary key information mining sub-network.

[0025] In some preferred embodiments, in the above-mentioned survey data processing method applied to image analysis, the arbitrary primary exemplary survey image is an arbitrary initial exemplary survey image or an exemplary adjustment survey image corresponding to the arbitrary initial exemplary survey image, and the exemplary adjustment survey image corresponding to the arbitrary initial exemplary survey image is a survey image formed by adjusting the image pixels in the arbitrary initial exemplary survey image;

[0026] The step of analyzing the network learning cost parameter corresponding to each primary exemplary survey image according to the image key information feature representation corresponding to each of the plurality of primary exemplary survey images comprises:

[0027] For the arbitrary initial exemplary survey image, the network learning cost parameter of the arbitrary initial exemplary survey image is analyzed according to the image key information feature representation of each initial exemplary survey image and the image key information feature representation of the exemplary adjustment survey image corresponding to the each initial exemplary survey image;

[0028] According to the image key information feature representation of each initial example mapping image and the image key information feature representation of the example adjustment mapping image corresponding to the initial example mapping image, a network learning cost parameter of the initial example mapping image is analyzed.

[0029] In some preferred embodiments, in the above-mentioned mapping data processing method applied to image analysis, the step of analyzing the network learning cost parameter of the initial example mapping image according to the image key information feature representation of each initial example mapping image and the image key information feature representation of the example adjustment mapping image corresponding to the initial example mapping image comprises:

[0030] According to the image key information feature representation of the initial example mapping image and the image key information feature representation of the example adjustment mapping image corresponding to the initial example mapping image, a first matching degree representation parameter between the initial example mapping image and the example adjustment mapping image corresponding to the initial example mapping image is calculated.

[0031] According to the image key information feature representation of the initial example mapping image and the image key information feature representation of other initial example mapping images, a second matching degree representation parameter between the initial example mapping image and the other initial example mapping images is calculated, the other initial example mapping images being initial example mapping images other than the initial example mapping image among the initial example mapping images.

[0032] According to the image key information feature representation of the initial example mapping image and the image key information feature representation of the example adjustment mapping image corresponding to the other initial example mapping image, a third matching degree representation parameter between the initial example mapping image and the example adjustment mapping image corresponding to the other initial example mapping image is calculated.

[0033] According to the first matching degree representation parameter, the second matching degree representation parameter and the third matching degree representation parameter, the network learning cost parameter of the initial example mapping image is calculated.

[0034] In some preferred embodiments, in the above-mentioned survey data processing method applied to image analysis, the step of analyzing the network learning cost parameter of the exemplary adjusted survey image corresponding to the arbitrary one of the initial exemplary survey images according to the image key information feature representation of the arbitrary one of the initial exemplary survey images and the image key information feature representation of the exemplary adjusted survey image corresponding to the arbitrary one of the initial exemplary survey images comprises:

[0035] calculating a first matching degree representation parameter between the arbitrary one of the initial exemplary survey images and the exemplary adjusted survey image corresponding to the arbitrary one of the initial exemplary survey images according to the image key information feature representation of the arbitrary one of the initial exemplary survey images and the image key information feature representation of the exemplary adjusted survey image corresponding to the arbitrary one of the initial exemplary survey images;

[0036] calculating a fourth matching degree representation parameter between the exemplary adjusted survey image corresponding to the arbitrary one of the initial exemplary survey images and other initial exemplary survey images according to the image key information feature representation of the exemplary adjusted survey image corresponding to the arbitrary one of the initial exemplary survey images and the image key information feature representation of the other initial exemplary survey images, the other initial exemplary survey images being initial exemplary survey images other than the arbitrary one of the initial exemplary survey images;

[0037] calculating a fifth matching degree representation parameter between the exemplary adjusted survey image corresponding to the arbitrary one of the initial exemplary survey images and the exemplary adjusted survey image corresponding to the other initial exemplary survey images according to the image key information feature representation of the exemplary adjusted survey image corresponding to the arbitrary one of the initial exemplary survey images and the image key information feature representation of the exemplary adjusted survey image corresponding to the other initial exemplary survey images;

[0038] calculating the network learning cost parameter of the exemplary adjusted survey image corresponding to the arbitrary one of the initial exemplary survey images according to the first matching degree representation parameter, the fourth matching degree representation parameter and the fifth matching degree representation parameter.

[0039] In some preferred embodiments, in the above-mentioned survey data processing method applied to image analysis, the step of performing network optimization operation on the primary key information mining sub-network according to the image key information feature representation corresponding to each of the plurality of primary exemplary survey images to form a senior key information mining sub-network corresponding to the primary key information mining sub-network comprises:

[0040] According to the image key information feature representation of each primary example mapping image, image pixel adjustment estimation data of each primary example mapping image is analyzed, and the image pixel adjustment estimation data of each primary example mapping image is a possibility parameter of each image pixel in the primary example mapping image being adjusted which is analyzed;

[0041] Image pixel adjustment actual data of each primary example mapping image is extracted, and the image pixel adjustment actual data of each primary example mapping image is data of whether each image pixel in the primary example mapping image is adjusted actually;

[0042] According to the image pixel adjustment estimation data of each primary example mapping image and the image pixel adjustment actual data of each primary example mapping image, network optimization operation is performed on the primary key information mining sub-network to form a high-level key information mining sub-network corresponding to the primary key information mining sub-network.

[0043] In some preferred embodiments, in the mapping data processing method applied to image analysis, the step of forming a corresponding optimized image analysis network according to the high-level key information mining sub-network comprises:

[0044] A high-level example mapping image is extracted, and image category identification data corresponding to the high-level example mapping image is extracted;

[0045] At least two fusion example mapping images are analyzed by using a high-level candidate neural network, the fusion example mapping image includes the high-level example mapping image and to-be-determined image category information, and the high-level candidate neural network includes the high-level key information mining sub-network;

[0046] According to the analysis possibility parameters of each image pixel in the at least two fusion example mapping images, a target fusion example mapping image is analyzed in the at least two fusion example mapping images;

[0047] According to the to-be-determined image category information in the target fusion example mapping image and the image category identification data of the high-level example mapping image, network optimization operation is performed on the high-level candidate neural network to form a corresponding optimized image analysis network.

[0048] The embodiment of the application also provides a mapping data processing system applied to image analysis, comprising a processor and a memory, the memory is used for storing a computer program, and the processor is used for executing the computer program to realize the mapping data processing method applied to image analysis.

[0049] The application embodiment provided in the application embodiment applied to image analysis survey data processing method and system can extract the survey image to be analyzed first; network optimization operation is performed to form an optimized image analysis network; the optimized image analysis network is used to analyze the analysis possibility parameters of each image pixel in at least two restored drawing images, the restored drawing images including the survey image to be analyzed and the pending image category information; the target restored drawing image is selected from the at least two restored drawing images according to the analysis possibility parameters of each image pixel in the at least two restored drawing images; and the pending image category information in the target restored drawing image is marked as the target image category information of the survey image to be analyzed. Based on the foregoing, the restored drawing image including the survey image to be analyzed and the pending image category information is directly analyzed and predicted, instead of only analyzing and predicting the survey image to be analyzed, so that the analysis prediction is more targeted, and therefore the reliability of the survey image analysis can be improved to some extent, thereby improving the deficiencies in the prior art.

[0050] In order to make the above object, characteristics and advantages of the present application more obvious and easy to understand, the following preferred embodiments are specifically described below, and the accompanying drawings are described in detail as follows. BRIEF DESCRIPTION OF DRAWINGS

[0051] Figure 1 The structural block diagram of the survey data processing system applied to image analysis provided by the application embodiment.

[0052] Figure 2 The flowchart of each step included in the survey data processing method applied to image analysis provided by the application embodiment.

[0053] Figure 3 The schematic diagram of each module included in the survey data processing device applied to image analysis provided by the application embodiment. EMBODIMENT

[0054] In order to make the object, technical scheme and advantages of the application embodiment more clear, the technical scheme in the application embodiment will be described clearly and completely in combination with the drawings in the application embodiment. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. The components of the application embodiments described and shown in the drawings can be arranged and designed in various different configurations.

[0055] Therefore, the following detailed description of the embodiments of the application provided in the drawings is not intended to limit the scope of the claimed application, but only represents selected embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those of ordinary skill in the art without creative labor are within the scope of protection of the application.

[0056] As Figure 1 shown, an embodiment of the present application provides a survey data processing system applied to image analysis. The survey data processing system applied to image analysis can include a memory and a processor.

[0057] In detail, the memory and the processor are directly or indirectly electrically connected to realize data transmission or interaction. For example, they can be electrically connected through one or more communication buses or signal lines. The memory can store at least one software function module (computer program) in the form of software or firmware. The processor can be used to execute the executable computer program stored in the memory, thereby realizing the survey data processing method applied to image analysis provided by the embodiment of the present application.

[0058] It can be understood that, in some feasible embodiments, the memory can be, but is not limited to, a random access memory (RAM), a read only memory (ROM), a programmable read only memory (PROM), an erasable programmable read only memory (EPROM), an electrically erasable programmable read only memory (EEPROM) and the like.

[0059] It can be understood that, in some feasible embodiments, the processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), a system on chip (SoC) and the like; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component.

[0060] It can be understood that, in some feasible embodiments, the survey data processing system applied to image analysis can be a server with data processing capability.

[0061] In combination Figure 2The embodiment of the present application also provides a surveying and mapping data processing method applied to image analysis, which can be applied to the surveying and mapping data processing system applied to image analysis. The method steps defined by the flow of the surveying and mapping data processing method applied to image analysis can be implemented by the surveying and mapping data processing system applied to image analysis.

[0062] The specific flow shown in FIG. 1 will be described in detail. Figure 2

[0063] In step S110, a surveying and mapping image to be analyzed is extracted.

[0064] In the embodiment of the present application, the surveying and mapping data processing system applied to image analysis can extract a surveying and mapping image to be analyzed. The surveying and mapping image to be analyzed is formed based on a target image surveying operation on a target region. For example, the surveying and mapping image to be analyzed is formed by an image acquisition operation of a target image acquisition device on the target region.

[0065] In step S120, an optimized image analysis network is formed by performing a network optimization operation.

[0066] In the embodiment of the present application, the surveying and mapping data processing system applied to image analysis can form an optimized image analysis network by performing a network optimization operation. For example, the network optimization operation can be performed on an original image analysis neural network to form an optimized image analysis network.

[0067] In step S130, an analysis possibility parameter of each image pixel in at least two restored mapping images is analyzed by using the optimized image analysis network.

[0068] In the embodiment of the present application, the surveying and mapping data processing system applied to image analysis can analyze an analysis possibility parameter of each image pixel in at least two restored mapping images by using the optimized image analysis network. The restored mapping images include the surveying and mapping image to be analyzed and pending image category information, that is, at least two kinds of pending image category information are fused into the surveying and mapping image to be analyzed to form at least two restored mapping images. In addition, the pending image category information can be text data. Therefore, when the fusion is performed, the text data can be converted into image data, and then the image data and the surveying and mapping image to be analyzed are spliced to form a restored mapping image. The image data can have characters corresponding to the text data in the image data, and other regions can be white regions. The analysis possibility parameter can be used to reflect the possibility of the appearance of a corresponding image pixel (for example, the possibility of a pixel value of a pixel point being a certain value).

[0069] ​Step S140, according to the analysis possibility parameter of each image pixel in the at least two restored mapping images, a target restored mapping image is selected from the at least two restored mapping images.

[0070] In the embodiments of the present application, the mapping data processing system applied to image analysis can select a target restored mapping image from the at least two restored mapping images according to the analysis possibility parameter of each image pixel in the at least two restored mapping images. For example, for each of the restored mapping images, the product of the analysis possibility parameters of each image pixel in the restored mapping image can be calculated, and then the restored mapping image corresponding to the product with the maximum value can be marked as the target restored mapping image, or selected in other ways.

[0071] Step S150, the pending image category information in the target restored mapping image is marked as the target image category information of the mapping image to be analyzed.

[0072] In the embodiments of the present application, the mapping data processing system applied to image analysis can mark the pending image category information in the target restored mapping image as the target image category information of the mapping image to be analyzed. The target image category information is used to reflect the image category of the mapping image to be analyzed, such as the abnormal category of the region, the feature category of the region, etc.

[0073] Based on the foregoing, since the restored mapping image including the mapping image to be analyzed and the pending image category information is directly analyzed and predicted, instead of only analyzing and predicting the mapping image to be analyzed, the analysis and prediction is more targeted, and therefore the reliability of the mapping image analysis can be improved to some extent, thereby improving the deficiencies in the prior art.

[0074] It can be understood that in some feasible embodiments, the step S120 described above, i.e., the step of forming an optimized image analysis network by performing network optimization, can further include the following implementation contents:

[0075] The plurality of primary exemplary mapping images are extracted, and the primary exemplary mapping images and the high-level exemplary mapping images described below are all exemplary mapping images (i.e., the basis for network optimization), and the primary and high-level are only used to distinguish each other, and do not have other special meanings;

[0076] For any one primary exemplary mapping image, the analysis possibility parameters of each image pixel in the any one primary exemplary mapping image are analyzed by using a primary candidate neural network. For example, for any one primary exemplary mapping image, the any one primary exemplary mapping image is loaded into the primary candidate neural network, and the image key information feature representation of the any one primary exemplary mapping image is determined by using the primary candidate neural network, and then the analysis possibility parameters of each image pixel in the any one primary exemplary mapping image are determined based on the image key information feature representation of the any one primary exemplary mapping image, wherein the image key information feature representation of the any one primary exemplary mapping image includes the image pixel key information feature representation of each image pixel in the any one primary exemplary mapping image. In addition, the analysis possibility parameter of any one image pixel in the primary exemplary mapping image can be the possibility of the appearance of the any one image pixel based on at least one image pixel before the any one image pixel.

[0077] According to the analysis possibility parameters of each image pixel in the plurality of primary exemplary mapping images, the primary candidate neural network is subjected to network optimization operation to form a primary optimized neural network, and the primary optimized neural network includes a primary key information mining subnetwork. For each of the primary exemplary mapping images, the ratio between the analysis possibility parameters of each image pixel in the primary exemplary mapping image and a preset parameter is calculated, and then the sum of the logarithmic operation results of each ratio is calculated to obtain the network optimization cost parameter corresponding to the primary exemplary mapping image. Based on the network optimization cost parameter, the network optimization operation is performed on the primary candidate neural network to form the primary optimized neural network, that is, the network parameters of the primary candidate neural network are updated and optimized, wherein the preset parameter belongs to the network parameters.

[0078] The image key information feature representation corresponding to the any one primary exemplary mapping image is mined by using the primary key information mining subnetwork. For example, the primary exemplary mapping image can be mapped to a feature space, and then the mapping result of the feature space can be subjected to linear processing to form the image key information feature representation, wherein the network parameters of the mapping to the feature space and the linear processing can be formed in the corresponding network optimization process.

[0079] According to the image key information feature representations corresponding to the plurality of primary exemplary mapping images, the primary key information mining subnetwork is subjected to network optimization operation to form a high-level key information mining subnetwork corresponding to the primary key information mining subnetwork.

[0080] According to the high-level key information mining subnetwork, a corresponding optimized image analysis network is formed.

[0081] It can be understood that, in some possible implementation manners, the step of performing network optimization operation on the primary key information mining sub-network according to the image key information feature representation corresponding to each of the plurality of primary exemplary mapping images to form a high-level key information mining sub-network corresponding to the primary key information mining sub-network can further include the following implementation manners:

[0082] According to the image key information feature representation corresponding to each of the plurality of primary exemplary mapping images, analyzing a network learning cost parameter corresponding to each of the plurality of primary exemplary mapping images;

[0083] According to the network learning cost parameter corresponding to each of the plurality of primary exemplary mapping images, analyzing a network learning cost parameter corresponding to the primary key information mining sub-network, for example, summing the network learning cost parameters corresponding to each of the plurality of primary exemplary mapping images to obtain the network learning cost parameter corresponding to the primary key information mining sub-network;

[0084] According to the network learning cost parameter corresponding to the primary key information mining sub-network, performing network optimization operation on the primary key information mining sub-network to form a high-level key information mining sub-network corresponding to the primary key information mining sub-network, for example, optimizing and updating the network parameters of the primary key information mining sub-network in a direction of reducing the network learning cost parameter corresponding to the primary key information mining sub-network.

[0085] It can be understood that, in some possible implementation manners, the any one primary exemplary mapping image is any one initial exemplary mapping image or an exemplary adjusted mapping image corresponding to the any one initial exemplary mapping image, the exemplary adjusted mapping image corresponding to the any one initial exemplary mapping image is a mapping image formed by adjusting image pixels in the any one initial exemplary mapping image (wherein the number of adjusted image pixels is not limited, such as at least one, when the number of adjusted image pixels is a plurality, the plurality of image pixels can be continuous or discontinuous, and the difference between the pixel value after adjustment and the pixel value before adjustment of one image pixel should be small, such as less than a preset pixel value, which can be configured according to actual needs), based on which, the step of analyzing a network learning cost parameter corresponding to each of the plurality of primary exemplary mapping images according to the image key information feature representation corresponding to each of the plurality of primary exemplary mapping images can further include the following implementation manners:

[0086] According to the image key information feature representation of each initial exemplified mapping image and the image key information feature representation of the exemplified adjustment mapping image corresponding to the initial exemplified mapping image, a network learning cost parameter of the initial exemplified mapping image is analyzed;

[0087] According to the image key information feature representation of each initial exemplified mapping image and the image key information feature representation of the exemplified adjustment mapping image corresponding to the initial exemplified mapping image, a network learning cost parameter of the initial exemplified mapping image is analyzed;

[0088] It can be understood that, in some possible embodiments, the step of analyzing the network learning cost parameter of the initial exemplified mapping image according to the image key information feature representation of each initial exemplified mapping image and the image key information feature representation of the exemplified adjustment mapping image corresponding to the initial exemplified mapping image can further include the following implementation contents:

[0089] According to the image key information feature representation of each initial exemplified mapping image and the image key information feature representation of the exemplified adjustment mapping image corresponding to the initial exemplified mapping image, a network learning cost parameter of the initial exemplified mapping image is analyzed;

[0090] According to the image key information feature representation of each initial exemplified mapping image and the image key information feature representation of the exemplified adjustment mapping image corresponding to the initial exemplified mapping image, a network learning cost parameter of the initial exemplified mapping image is analyzed;

[0091] According to the image key information feature representation of each initial exemplified mapping image and the image key information feature representation of the exemplified adjustment mapping image corresponding to the initial exemplified mapping image, a network learning cost parameter of the initial exemplified mapping image is analyzed;

[0092] According to the first matching degree representation parameter, the second matching degree representation parameter and the third matching degree representation parameter, a network learning cost parameter of the arbitrary initial example mapping image is calculated.

[0093] It can be understood that, in some possible embodiments, the step of calculating the first matching degree representation parameter between the arbitrary initial example mapping image and the example adjustment mapping image corresponding to the arbitrary initial example mapping image according to the image key information feature representation of the arbitrary initial example mapping image and the image key information feature representation of the example adjustment mapping image corresponding to the arbitrary initial example mapping image can further include the following implementation contents:

[0094] The image key information feature representation of the arbitrary initial example mapping image and the image key information feature representation of the example adjustment mapping image corresponding to the arbitrary initial example mapping image are calculated by a number product to output a corresponding feature representation number product, and based on a target adjustment parameter, the feature representation number product is adjusted to form the first matching degree representation parameter between the arbitrary initial example mapping image and the example adjustment mapping image corresponding to the arbitrary initial example mapping image. For example, the first matching degree representation parameter can have a positive correlation with the feature representation number product, and the second matching degree representation parameter can have a negative correlation with the target adjustment parameter. The target adjustment parameter can be determined in the following manner: a maximum network optimization stage number (which can be preconfigured, i.e., a maximum number of network optimization, a stage number or a round number, etc.) is obtained, and a current network optimization stage number is obtained, an absolute difference value between a positive correlation value of the current network optimization stage number and the maximum network optimization stage number (a ratio between the positive correlation value and the maximum network optimization stage number is equal to a fixed value, such as 0.5) is calculated, a ratio between the absolute difference value and the maximum network optimization stage number is calculated, and finally, based on a pre-determined offset parameter, the ratio is updated (such as addition), to obtain the target adjustment parameter. The offset parameter can be 0.01, 0.02, 0.03, 0.04, etc.

[0095] It can be understood that, in some possible embodiments, the step of calculating the network learning cost parameter of the arbitrary initial example mapping image according to the first matching degree representation parameter, the second matching degree representation parameter and the third matching degree representation parameter can further include the following implementation contents:

[0096] calculating an exponential operation result of the first matching degree representation parameter to obtain a first exponential operation result, calculating an exponential operation result of the second matching degree representation parameter to obtain a second exponential operation result, and calculating an exponential operation result of the third matching degree representation parameter to obtain a third exponential operation result;

[0097] calculating a sum value of each second exponential operation result and each third exponential operation result to obtain a target exponential operation result, calculating a ratio between the first exponential operation result and the target exponential operation result, and performing a logarithm operation on the ratio, and finally determining the network learning cost parameter of the arbitrary initial exemplary mapping image based on a result of the logarithm operation, for example, the network learning cost parameter can be negatively correlated with the result of the logarithm operation, and in another embodiment, a sum value of the second matching degree representation parameter and the third matching degree representation parameter can be directly calculated, a ratio between the first matching degree representation parameter and the sum value can be calculated, and a logarithm operation is performed on the ratio, and finally the network learning cost parameter of the arbitrary initial exemplary mapping image is determined based on a result of the logarithm operation.

[0098] It can be understood that in some possible embodiments, the step of analyzing the network learning cost parameter of the exemplary adjusted mapping image corresponding to the arbitrary initial exemplary mapping image according to the image key information feature representation of the arbitrary initial exemplary mapping image and the image key information feature representation of the exemplary adjusted mapping image corresponding to the arbitrary initial exemplary mapping image can further include the following implementation contents:

[0099] calculating a first matching degree representation parameter between the arbitrary initial exemplary mapping image and the exemplary adjusted mapping image corresponding to the arbitrary initial exemplary mapping image according to the image key information feature representation of the arbitrary initial exemplary mapping image and the image key information feature representation of the exemplary adjusted mapping image corresponding to the arbitrary initial exemplary mapping image, as described above;

[0100] calculating a fourth matching degree representation parameter between the exemplary adjusted mapping image corresponding to the arbitrary initial exemplary mapping image and other initial exemplary mapping images according to the image key information feature representation of the exemplary adjusted mapping image corresponding to the arbitrary initial exemplary mapping image and the image key information feature representation of the other initial exemplary mapping images, the other initial exemplary mapping images being initial exemplary mapping images other than the arbitrary initial exemplary mapping image among the initial exemplary mapping images, as described above.

[0101] According to the image key information feature representation of the exemplary adjusted mapping image corresponding to the arbitrary one of the initial exemplary mapping images and the image key information feature representation of the exemplary adjusted mapping image corresponding to the other initial exemplary mapping images, a fifth matching degree representation parameter between the exemplary adjusted mapping image corresponding to the arbitrary one of the initial exemplary mapping images and the exemplary adjusted mapping image corresponding to the other initial exemplary mapping images is calculated, as previously described.

[0102] According to the first matching degree representation parameter, the fourth matching degree representation parameter and the fifth matching degree representation parameter, a network learning cost parameter of the exemplary adjusted mapping image corresponding to the arbitrary one of the initial exemplary mapping images is calculated, as previously described.

[0103] It can be understood that, in some possible implementation manners, the step of performing network optimization operation on the primary key information mining sub-network according to the image key information feature representation corresponding to each of the plurality of primary exemplary mapping images to form a senior key information mining sub-network corresponding to the primary key information mining sub-network can further include the following implementation contents:

[0104] According to the image key information feature representation of each primary exemplary mapping image, image pixel adjustment estimation data of each primary exemplary mapping image is analyzed, the image pixel adjustment estimation data of the primary exemplary mapping image being a possibility parameter of each image pixel in the primary exemplary mapping image being adjusted which is analyzed;

[0105] Image pixel adjustment actual data of each primary exemplary mapping image is extracted, the image pixel adjustment actual data of the primary exemplary mapping image being data of whether each image pixel in the primary exemplary mapping image is adjusted actually;

[0106] According to the image pixel adjustment estimation data of each primary exemplary mapping image and the image pixel adjustment actual data of each primary exemplary mapping image, network optimization operation is performed on the primary key information mining sub-network to form a senior key information mining sub-network corresponding to the primary key information mining sub-network, that is, a corresponding network optimization cost parameter can be determined based on the difference between the image pixel adjustment estimation data and the image pixel adjustment actual data, and then network optimization operation is performed on the primary key information mining sub-network based on the network optimization cost parameter to form a senior key information mining sub-network.

[0107] It can be understood that, in some possible implementation manners, the step of forming the corresponding optimized image analysis network according to the high-level key information mining sub-network can further include the following implementation manners:

[0108] The high-level exemplary mapping image is extracted, and image category identification data corresponding to the high-level exemplary mapping image is extracted, and the image category identification data can be formed by manual annotation;

[0109] The analysis possibility parameters of each image pixel in at least two fusion exemplary mapping images are analyzed by using a high-level candidate neural network, the fusion exemplary mapping images include the high-level exemplary mapping image and to-be-determined image category information, and the high-level candidate neural network includes the high-level key information mining sub-network;

[0110] The target fusion exemplary mapping image is analyzed from the at least two fusion exemplary mapping images according to the analysis possibility parameters of each image pixel in the at least two fusion exemplary mapping images;

[0111] The high-level candidate neural network is subjected to network optimization operation to form a corresponding optimized image analysis network according to the to-be-determined image category information in the target fusion exemplary mapping image and the image category identification data of the high-level exemplary mapping image, for example, a corresponding network optimization cost parameter can be determined according to a difference between the to-be-determined image category information and the image category identification data, and the high-level candidate neural network is subjected to network optimization operation to form the corresponding optimized image analysis network based on the network optimization cost parameter.

[0112] It can be understood that, in some possible implementation manners, the step of analyzing the analysis possibility parameters of each image pixel in the at least two fusion exemplary mapping images by using the high-level candidate neural network can further include the following implementation manners:

[0113] The data fusion operation is performed on the high-level exemplary mapping image and the to-be-determined image category information to form one fusion exemplary mapping image in the at least two fusion exemplary mapping images, as described above.

[0114] The one fusion exemplary mapping image is loaded into the high-level candidate neural network, the key information feature representation of the one fusion exemplary mapping image is mined by using the high-level candidate neural network, and the analysis possibility parameters of each image pixel in the one fusion exemplary mapping image are analyzed according to the key information feature representation of the one fusion exemplary mapping image, as described above.

[0115] In some possible implementation manners, the step of analyzing the analysis possibility parameters of the pixels in each of the at least two fused exemplary mapping images by using the high-level candidate neural network can further include the following implementation manners:

[0116] loading the high-level exemplary mapping image into the high-level candidate neural network, and mining the key information feature representation of the high-level exemplary mapping image by using the high-level candidate neural network, as described above;

[0117] performing an aggregation operation on the key information feature representation of the high-level exemplary mapping image and the key information feature representation of the to-be-determined image category information by using the high-level candidate neural network, to form the key information feature representation of one of the at least two fused exemplary mapping images, as described above;

[0118] analyzing the analysis possibility parameters of the pixels in the one of the at least two fused exemplary mapping images by using the high-level candidate neural network according to the key information feature representation of the one of the at least two fused exemplary mapping images, as described above.

[0119] In some possible implementation manners, the step of analyzing the target fused exemplary mapping image from the at least two fused exemplary mapping images according to the analysis possibility parameters of the pixels in the at least two fused exemplary mapping images can further include the following implementation manners:

[0120] analyzing, according to the analysis possibility parameters of the pixels in the at least two fused exemplary mapping images, an image content connection parameter of each of the at least two fused exemplary mapping images, the image content connection parameter being used to reflect a connection degree (i.e., a matching degree) between image contents included in the fused exemplary mapping image;

[0121] screening, according to the image content connection parameters of the at least two fused exemplary mapping images, a fused exemplary mapping image whose corresponding image content connection parameter matches a preset connection parameter, and marking the fused exemplary mapping image as the target fused exemplary mapping image, for example, a fused exemplary mapping image whose image content connection parameter has a maximum value can be determined as the fused exemplary mapping image whose corresponding image content connection parameter matches the preset connection parameter, and thus the fused exemplary mapping image can be marked as the target fused exemplary mapping image.

[0122] In some possible implementation, the step of analyzing the image content continuity parameter of each of the fused sample mapping images according to the analysis possibility parameter of each image pixel in the fused sample mapping image, can further include the following implementation.

[0123] analyzing the analysis possibility parameter of each of the fused sample mapping images according to the analysis possibility parameter of each image pixel in the fused sample mapping image;

[0124] analyzing the image content continuity parameter of each of the fused sample mapping images according to the analysis possibility parameter of the fused sample mapping image.

[0125] In some possible implementation, the step of analyzing the analysis possibility parameter of each of the fused sample mapping images according to the analysis possibility parameter of each image pixel in the fused sample mapping image, can further include the following implementation.

[0126] For each of the fused sample mapping images, the analysis possibility parameter of each image pixel in the fused sample mapping image is fused, such as multiplied, to output the analysis possibility parameter of the fused sample mapping image.

[0127] In some possible implementation, the step of analyzing the image content continuity parameter of each of the fused sample mapping images according to the analysis possibility parameter of each of the fused sample mapping images, can further include the following implementation.

[0128] For each of the fused sample mapping images, the number of image pixels included in the fused sample mapping image is determined, and a first negative correlation parameter of the analysis possibility parameter of the fused sample mapping image (for example, the product of the first negative correlation parameter and the analysis possibility parameter is equal to a fixed value, such as 0.5, 1, 2) and a second negative correlation parameter of the number of image pixels (for example, the product of the second negative correlation parameter and the number of image pixels is equal to a fixed value, such as 0.5, 1, 2) are calculated, and the first negative correlation parameter is operated by power operation based on the second negative correlation parameter (for example, the second negative correlation parameter is operated by the first negative correlation parameter corresponding times of power operation) to output the image content continuity parameter of the fused sample mapping image. For example, the image content continuity parameter can have a negative correlation corresponding relationship with the result of the power operation, such as the product of a fixed value.

[0129] It can be understood that, in some feasible embodiments, the step S130 in the foregoing description, that is, the step of analyzing the analysis possibility parameters of each image pixel in the at least two restored mapping images by using the optimized image analysis network, can further include the following implementation contents:

[0130] performing a data fusion operation on the to-be-analyzed mapping image and the to-be-determined image category information to form one of the at least two restored mapping images, as described above;

[0131] loading the one restored mapping image into the optimized image analysis network, mining a key information feature representation of the one restored mapping image by using the optimized image analysis network, and analyzing the analysis possibility parameters of each image pixel in the one restored mapping image according to the key information feature representation of the one restored mapping image, for example, the restored mapping image can be encoded by the optimized image analysis network to represent the restored mapping image in the form of a vector, that is, to obtain a corresponding key information feature representation, then, the key information feature representation can be subjected to a full connection operation to form a key information full connection feature representation, and finally, the key information full connection feature representation can be subjected to an activation operation to form the analysis possibility parameters of each image pixel in the one restored mapping image, which can be implemented by using an activation function, which can be a softmax function or the like.

[0132] It can be understood that, in some feasible embodiments, the step S130 in the foregoing description, that is, the step of analyzing the analysis possibility parameters of each image pixel in the at least two restored mapping images by using the optimized image analysis network, can further include the following implementation contents:

[0133] loading the to-be-analyzed mapping image into the optimized image analysis network to mine a key information feature representation of the to-be-analyzed mapping image, and performing an aggregation operation, such as a cascade combination operation, on the key information feature representation of the to-be-analyzed mapping image and the key information feature representation of the to-be-determined image category information by using the optimized image analysis network to form a key information feature representation of one of the at least two restored mapping images, that is, the to-be-analyzed mapping image and the to-be-determined image category information can be subjected to feature mining respectively to form corresponding key information feature representations, and the analysis possibility parameters of each image pixel in the one restored mapping image can be analyzed according to the key information feature representation of the one restored mapping image by using the optimized image analysis network, as described above.

[0134] In combination withFigure 3 The embodiment of the present application also provides a surveying and mapping data processing device applied to image analysis, which can be applied to the surveying and mapping data processing system applied to image analysis.

[0135] The surveying and mapping image extraction module is configured to extract a to-be-analyzed surveying and mapping image, wherein the to-be-analyzed surveying and mapping image is formed based on a target image surveying operation performed on a target region.

[0136] The neural network optimization module is configured to form an optimized image analysis network by performing a network optimization operation.

[0137] The surveying and mapping image analysis module is configured to analyze an analysis possibility parameter of each image pixel in at least two restored mapping images by using the optimized image analysis network, wherein the at least two restored mapping images include the to-be-analyzed surveying and mapping image and pending image category information.

[0138] The surveying and mapping image determination module is configured to select a target restored mapping image from the at least two restored mapping images according to the analysis possibility parameter of each image pixel in the at least two restored mapping images.

[0139] The image category information determination module is configured to mark the pending image category information in the target restored mapping image as target image category information of the to-be-analyzed surveying and mapping image, wherein the target image category information is used to reflect the image category of the to-be-analyzed surveying and mapping image.

[0140] In summary, the surveying and mapping data processing method and system applied to image analysis provided by the present application can extract a to-be-analyzed surveying and mapping image, form an optimized image analysis network by performing a network optimization operation, analyze an analysis possibility parameter of each image pixel in at least two restored mapping images by using the optimized image analysis network, select a target restored mapping image from the at least two restored mapping images according to the analysis possibility parameter of each image pixel in the at least two restored mapping images, and mark the pending image category information in the target restored mapping image as target image category information of the to-be-analyzed surveying and mapping image. Based on the foregoing, the restored mapping image including the to-be-analyzed surveying and mapping image and the pending image category information is directly analyzed and predicted, rather than only the to-be-analyzed surveying and mapping image is analyzed and predicted, so that the analysis and prediction is more targeted, and therefore, the reliability of surveying and mapping image analysis can be improved to some extent, thereby improving the deficiencies in the prior art.

[0141] The above merely provides the preferred embodiments of the present application, and is not used to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modifications, equivalent replacements, improvements, etc. made within the principles and technical scope of the present application shall fall into the scope of the present application.

Claims

1. A method for processing survey data applied to image analysis, characterized by, The method comprises the following steps: extracting a to-be-analyzed mapping image, which is based on a target image mapping operation performed on a target region to form a target image category information; forming an optimized image analysis network by performing network optimization operations; analyzing, by using the optimized image analysis network, analysis possibility parameters of each image pixel in at least two restored mapping images, the restored mapping images including the to-be-analyzed mapping image and to-be-determined image category information; selecting, according to the analysis possibility parameters of each image pixel in the at least two restored mapping images, a target restored mapping image from the at least two restored mapping images; labeling the to-be-determined image category information in the target restored mapping image as the target image category information of the to-be-analyzed mapping image, the target image category information being used to reflect the image category of the to-be-analyzed mapping image; the step of forming an optimized image analysis network by performing network optimization operations comprises the following steps: extracting a plurality of primary exemplary mapping images; for any one of the primary exemplary mapping images, analyzing, by using a primary candidate neural network, analysis possibility parameters of each image pixel in the any one of the primary exemplary mapping images; performing network optimization operations on the primary candidate neural network according to the analysis possibility parameters of each image pixel in the plurality of primary exemplary mapping images to form a primary optimized neural network, the primary optimized neural network including a primary key information mining sub-network; mining, by using the primary key information mining sub-network, image key information feature representations corresponding to the any one of the primary exemplary mapping images; performing network optimization operations on the primary key information mining sub-network according to the image key information feature representations corresponding to the plurality of primary exemplary mapping images to form a senior key information mining sub-network corresponding to the primary key information mining sub-network; forming a corresponding optimized image analysis network according to the senior key information mining sub-network.

2. The plot data processing method for image analysis according to claim 1, wherein, the step of analyzing, by using the optimized image analysis network, analysis possibility parameters of each image pixel in at least two restored mapping images comprises: performing data fusion operations on the to-be-analyzed mapping image and the to-be-determined image category information to form one of the at least two restored mapping images; and loading the one of the restored mapping images into the optimized image analysis network, mining, by using the optimized image analysis network, key information feature representations of the one of the restored mapping images, and analyzing, according to the key information feature representations of the one of the restored mapping images, analysis possibility parameters of each image pixel in the one of the restored mapping images; or loading the to-be-analyzed mapping image into the optimized image analysis network to mine key information feature representation of the to-be-analyzed mapping image by using the optimized image analysis network, and performing an aggregation operation on the key information feature representation of the to-be-analyzed mapping image and the key information feature representation of the to-be-determined image category information by using the optimized image analysis network to form key information feature representation of one of the at least two restored mapping images, and analyzing an analysis possibility parameter of each image pixel in the one restored mapping image according to the key information feature representation of the one restored mapping image by using the optimized image analysis network.

3. The plot data processing method for image analysis according to claim 1, wherein, The step of performing network optimization operation on the primary key information mining sub-network according to the network learning cost parameter corresponding to each primary exemplary mapping image to form a high-level key information mining sub-network corresponding to the primary key information mining sub-network comprises: analyzing a network learning cost parameter corresponding to each primary exemplary mapping image according to the image key information feature representation corresponding to each primary exemplary mapping image; analyzing a network learning cost parameter corresponding to the primary key information mining sub-network according to the network learning cost parameters corresponding to each primary exemplary mapping image; performing network optimization operation on the primary key information mining sub-network according to the network learning cost parameter corresponding to the primary key information mining sub-network to form a high-level key information mining sub-network corresponding to the primary key information mining sub-network.

4. The plot data processing method for image analysis according to claim 3, wherein, The any one initial exemplary mapping image corresponds to an exemplary adjusted mapping image of the any one initial exemplary mapping image, and the exemplary adjusted mapping image of the any one initial exemplary mapping image is a mapping image formed by adjusting image pixels in the any one initial exemplary mapping image; The step of analyzing a network learning cost parameter corresponding to each primary exemplary mapping image according to the image key information feature representation corresponding to each primary exemplary mapping image comprises: for the any one initial exemplary mapping image, analyzing a network learning cost parameter of the any one initial exemplary mapping image according to the image key information feature representation of each initial exemplary mapping image and the image key information feature representation of the exemplary adjusted mapping image corresponding to each initial exemplary mapping image; for the exemplary adjusted mapping image corresponding to the any one initial exemplary mapping image, analyzing a network learning cost parameter of the exemplary adjusted mapping image corresponding to the any one initial exemplary mapping image according to the image key information feature representation of each initial exemplary mapping image and the image key information feature representation of the exemplary adjusted mapping image corresponding to each initial exemplary mapping image.

5. The plot data processing method for image analysis according to claim 4, wherein, The step of analyzing the network learning cost parameter of the arbitrary one initial example mapping image according to the image key information feature representation of each initial example mapping image and the image key information feature representation of the example adjustment mapping image corresponding to the initial example mapping image, comprises: According to the image key information feature representation of the arbitrary one initial example mapping image and the image key information feature representation of the example adjustment mapping image corresponding to the initial example mapping image, the first matching degree representation parameter between the arbitrary one initial example mapping image and the example adjustment mapping image corresponding to the initial example mapping image is calculated; According to the image key information feature representation of the arbitrary one initial example mapping image and the image key information feature representation of the other initial example mapping image, the second matching degree representation parameter between the arbitrary one initial example mapping image and the other initial example mapping image is calculated, and the other initial example mapping image is the initial example mapping image other than the arbitrary one initial example mapping image in the initial example mapping image. According to the image key information feature representation of the arbitrary one initial example mapping image and the image key information feature representation of the example adjustment mapping image corresponding to the other initial example mapping image, the third matching degree representation parameter between the arbitrary one initial example mapping image and the example adjustment mapping image corresponding to the other initial example mapping image is calculated. According to the first matching degree representation parameter, the second matching degree representation parameter and the third matching degree representation parameter, the network learning cost parameter of the arbitrary one initial example mapping image is calculated.

6. The plot data processing method for image analysis according to claim 4, wherein, The step of analyzing the network learning cost parameter of the example adjustment mapping image corresponding to the arbitrary one initial example mapping image according to the image key information feature representation of each initial example mapping image and the image key information feature representation of the example adjustment mapping image corresponding to each initial example mapping image, comprises: According to the image key information feature representation of the arbitrary one initial example mapping image and the image key information feature representation of the example adjustment mapping image corresponding to the initial example mapping image, the first matching degree representation parameter between the arbitrary one initial example mapping image and the example adjustment mapping image corresponding to the initial example mapping image is calculated; According to the image key information feature representation of the arbitrary one initial example mapping image and the image key information feature representation of the example adjustment mapping image corresponding to the initial example mapping image, the first matching degree representation parameter between the arbitrary one initial example mapping image and the example adjustment mapping image corresponding to the initial example mapping image is calculated; According to the image key information feature representation of the exemplary adjusted mapping image corresponding to the arbitrary one of the initial exemplary mapping images and the image key information feature representation of the other initial exemplary mapping images, a fourth matching degree representation parameter between the exemplary adjusted mapping image corresponding to the arbitrary one of the initial exemplary mapping images and the other initial exemplary mapping images is calculated, the other initial exemplary mapping images belong to the initial exemplary mapping images other than the arbitrary one of the initial exemplary mapping images; According to the image key information feature representation of the exemplary adjusted mapping image corresponding to the arbitrary one of the initial exemplary mapping images and the image key information feature representation of the other initial exemplary mapping images, a fifth matching degree representation parameter between the exemplary adjusted mapping image corresponding to the arbitrary one of the initial exemplary mapping images and the other initial exemplary mapping images is calculated; According to the first matching degree representation parameter, the fourth matching degree representation parameter and the fifth matching degree representation parameter, a network learning cost parameter of the exemplary adjusted mapping image corresponding to the arbitrary one of the initial exemplary mapping images is calculated.

7. The plot data processing method for image analysis according to claim 1, wherein, The step of performing network optimization operation on the primary key information mining sub-network according to the image key information feature representation of each primary exemplary mapping image to form a senior key information mining sub-network corresponding to the primary key information mining sub-network, comprises: According to the image key information feature representation of each primary exemplary mapping image, image pixel adjustment estimation data of each primary exemplary mapping image is analyzed, the image pixel adjustment estimation data of the primary exemplary mapping image is a possibility parameter of each image pixel in the primary exemplary mapping image being adjusted; Image pixel adjustment actual data of each primary exemplary mapping image is extracted, the image pixel adjustment actual data of the primary exemplary mapping image is data of whether each image pixel in the primary exemplary mapping image is adjusted; According to the image pixel adjustment estimation data of each primary exemplary mapping image and the image pixel adjustment actual data of each primary exemplary mapping image, network optimization operation is performed on the primary key information mining sub-network to form a senior key information mining sub-network corresponding to the primary key information mining sub-network.

8. The plot data processing method for image analysis according to claim 1, wherein, The step of forming a corresponding optimized image analysis network according to the senior key information mining sub-network, comprises: A senior exemplary mapping image is extracted, and image category identification data corresponding to the senior exemplary mapping image is extracted; Using a senior candidate neural network, analysis possibility parameters of each image pixel in at least two fusion exemplary mapping images are analyzed, the fusion exemplary mapping images include the senior exemplary mapping image and undetermined image category information, and the senior candidate neural network includes the senior key information mining sub-network; According to the analysis possibility parameters of each image pixel in the at least two fusion example mapping images, a target fusion example mapping image is analyzed from the at least two fusion example mapping images; According to the to-be-determined image category information in the target fusion example mapping image and the image category identification data of the high-level example mapping image, the high-level candidate neural network is subjected to network optimization operation to form a corresponding optimized image analysis network.

9. A mapping data processing system for use in image analysis, characterized by A computer program product is provided, comprising a computer readable medium, having thereon computer program code, which, when executed by a computer, causes the computer to carry out the steps of the method according to any one of claims 1-8.

Images