Data processing method and device, and urban terrain classification method and device

Abstract

The embodiment of the invention provides a data processing method and device and an urban ground feature classification method and device, and the data processing method comprises the steps: collecting hyperspectral region data for a target region, and extracting the features of the target region from the hyperspectral region data; the target area features are input into an object classification model to classify area objects in the target area, object classification information of the area objects in the target area is obtained, and the object classification model optimizes the recognition precision of the object classification model for a mixed pixel area through a layered sensitive spectrum loss function; comparing the hyperspectral region data with historical hyperspectral region data of the target region to obtain region change information of the target region; and constructing a visual area report corresponding to the target area based on the object classification information and the area change information.

Description

Technical Field

[0001] The embodiments of the present specification relate to the field of data processing technology, and in particular, to a data processing method and device, and an urban feature classification method and device. Background Art

[0002] Classification and dynamic monitoring of urban objects are core tasks in urban planning and management. They aim to accurately obtain the spatial distribution and change information of objects in the city, and provide support for decision-making such as land use planning, greening supervision and building distribution optimization. However, traditional monitoring methods, such as satellite remote sensing and manual verification, are inefficient in processing large-scale and complex data. In urban environments with high building density and frequent changes in green space coverage, the monitoring accuracy is insufficient and is easily affected by environmental factors such as light and weather, which cannot meet the dynamic monitoring needs of the rapid development of the city. Monitoring objects through drones has become an optional solution. Its advantages of flexible deployment, efficient collection and low cost make it particularly suitable for urban environmental monitoring, especially when equipped with hyperspectral sensors, it can provide more accurate information about objects. However, after the drones in the prior art collect spectral data through hyperspectral sensors, although they can meet the requirements of object classification, in scenarios with high environmental complexity, the accuracy of the algorithm cannot meet the requirements, and there are certain errors in the classification of objects. Therefore, an effective solution is urgently needed to solve the above problems. Summary of the invention

[0003] In view of this, an embodiment of this specification provides a data processing method. One or more embodiments of this specification also relate to a method for classifying urban features, a data processing device, an urban feature classification device, a computing device, a computer-readable storage medium, and a computer program product to solve the technical defects existing in the prior art.

[0004] According to a first aspect of an embodiment of this specification, a data processing method is provided, including: Collecting hyperspectral region data for a target region, and extracting target region features from the hyperspectral region data; Inputting the target area feature into the object classification model to classify the area object in the target area, and obtaining the object classification information of the area object in the target area, wherein the object classification model optimizes the recognition accuracy of the object classification model for the mixed pixel area through a hierarchical sensitive spectral loss function; Comparing the hyperspectral region data with historical hyperspectral region data of the target region to obtain regional change information of the target region; A visual area report corresponding to the target area is constructed based on the object classification information and the area change information.

[0005] According to a second aspect of an embodiment of this specification, a method for classifying urban features is provided, comprising: Collecting hyperspectral regional data for urban areas, and extracting urban area features from the hyperspectral regional data; Inputting the urban area features into a classification model to classify the ground objects in the urban area, and obtaining ground object classification information of the ground objects in the urban area, wherein the classification model optimizes the recognition accuracy of the classification model for mixed pixel areas through a hierarchical sensitive spectral loss function; Comparing the hyperspectral region data with historical hyperspectral region data of the urban area to obtain regional change information of the urban area; An urban area change report corresponding to the urban area is constructed based on the land feature classification information and the area change information.

[0006] According to a third aspect of an embodiment of this specification, there is provided a data processing device, including: A collection module is configured to collect high-spectral region data for a target region and extract target region features from the high-spectral region data; An input module is configured to input the target area feature into an object classification model to classify the area object in the target area, and obtain object classification information of the area object in the target area, wherein the object classification model optimizes the recognition accuracy of the object classification model for the mixed pixel area through a hierarchical sensitive spectral loss function; A comparison module is configured to compare the hyperspectral region data with historical hyperspectral region data of the target region to obtain region change information of the target region; A construction module is configured to construct a visual area report corresponding to the target area based on the object classification information and the area change information.

[0007] According to a fourth aspect of the embodiments of this specification, there is provided an urban feature classification device, comprising: A data collection module is configured to collect high-spectral area data for urban areas and extract urban area features from the high-spectral area data; An input model module is configured to input the urban area features into a classification model to classify the ground objects in the urban area, and obtain ground object classification information of the ground objects in the urban area, wherein the classification model optimizes the recognition accuracy of the classification model for mixed pixel areas through a hierarchical sensitive spectral loss function; A data comparison module is configured to compare the hyperspectral region data with historical hyperspectral region data of the urban region to obtain regional change information of the urban region; A report building module is configured to build an urban area change report corresponding to the urban area based on the land feature classification information and the area change information.

[0008] According to a fifth aspect of an embodiment of this specification, a computing device is provided, including: Memory and processor; The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions. When the computer executable instructions are executed by the processor, the steps of the above-mentioned data processing method or urban feature classification method are implemented.

[0009] According to a sixth aspect of the embodiments of this specification, a computer-readable storage medium is provided, which stores computer-executable instructions, and when the instructions are executed by a processor, the steps of the above-mentioned data processing method or urban feature classification method are implemented.

[0010] According to a seventh aspect of the embodiments of this specification, a computer program product is provided, including a computer program or instructions, which, when executed by a processor, implement the steps of the above-mentioned data processing method or urban feature classification method.

[0011] The data processing method provided in this embodiment, in order to improve the accuracy of ground object recognition and support efficient and accurate regional planning, can first collect hyperspectral regional data for the target area, and extract the target area features from the hyperspectral regional data; on this basis, in order to improve the efficiency of object classification, the target area features can be input into the object classification model to classify the regional objects in the target area, so as to obtain the object classification information of the regional objects in the target area according to the classification results; in order to enable the classification model to have a strong object boundary recognition ability, the recognition accuracy of the object classification model for the mixed pixel area can be optimized by the hierarchical sensitive spectral loss function in the training stage; so that the optimized object classification model can still have a strong boundary recognition accuracy for the data collected in the complex environment, so as to ensure that the object classification information of the regional objects in the target area output by the model is more accurate. Thereafter, in order to facilitate the use for regional planning and management, the hyperspectral regional data can be compared with the historical hyperspectral regional data of the target area to obtain the regional change information of the target area; and then the visual regional report corresponding to the target area can be constructed based on the object classification information and the regional change information. When the ground object analysis and processing is realized, the accuracy and efficiency can be effectively improved, so as to facilitate the use of downstream business for regional planning and management. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 is a flow chart of a data processing method provided by an embodiment of this specification; Figure 2is a flow chart of a method for classifying urban features provided by an embodiment of this specification; Figure 3 is a processing flow chart of a data processing method provided by an embodiment of this specification; Figure 4 is a structural schematic diagram of a data processing device provided by an embodiment of this specification; Figure 5 It is a structural schematic diagram of a device for classifying urban features provided by an embodiment of this specification; Figure 6 It is a structural block diagram of a computing device provided by an embodiment of this specification. DETAILED DESCRIPTION

[0013] Many specific details are described in the following description to facilitate a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of this specification, so this specification is not limited to the specific implementation disclosed below.

[0014] The terms used in one or more embodiments of this specification are only for the purpose of describing specific embodiments, and are not intended to limit one or more embodiments of this specification. The singular forms of "a", "said" and "the" used in one or more embodiments of this specification and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0015] It should be understood that although the terms first, second, etc. may be used to describe various information in one or more embodiments of this specification, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of one or more embodiments of this specification, the first may also be referred to as the second, and similarly, the second may also be referred to as the first. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0016] In addition, it should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in one or more embodiments of this specification are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0017] First, the terms involved in one or more embodiments of this specification are explained.

[0018] Convolutional Neural Network (CNN) is a deep learning algorithm or model that is particularly suitable for processing data with a grid-like topological structure, such as image data (which can be viewed as a two-dimensional pixel grid).

[0019] Hyperspectral region data (also known as imaging spectral data) is a multi-dimensional information collection. It adds a spectral dimension to the traditional two-dimensional image data (spatial dimension and spectral dimension), forming a three-dimensional cube data structure.

[0020] Urban feature classification refers to the classification of various natural and artificial objects on the urban surface according to their formation process, nature, purpose, size and other characteristics. This classification helps to better understand the structure and characteristics of urban geographical space and further infer the urban environment and human activities.

[0021] Gaussian filter is a low-pass filter that is widely used in fields such as image processing. It can smooth images, reduce noise, and retain the edge and detail information of the image.

[0022] The Z-score standardization method is a common data preprocessing technique that converts raw data into deviations from the mean in units of standard deviation. The specific formula is: Z = (X - μ) / σ, where X is the original data point, μ is the mean of the data set, and σ is the standard deviation of the data set. The mean of the standardized data is 0 and the variance is 1, so that data of different magnitudes can be compared and analyzed on the same scale. This method plays an important role in data analysis and statistical modeling. It can eliminate the influence of different data units and data magnitudes, and improve the accuracy and reliability of data analysis and statistical modeling. At the same time, Z-score standardization also helps to identify outliers in the data set, providing an important basis for data cleaning and preprocessing.

[0023] Euclidean distance, also known as Euclidean distance, is the "normal" (i.e. straight-line) distance between two points in Euclidean space. In two-dimensional space, the formula for Euclidean distance is d = sqrt((x1-x2)^2+(y1-y2)^2); in three-dimensional space, the formula is d=sqrt((x1-x2)^2+(y1-y2)^2+(z1-z2)^2). Euclidean distance is a commonly used distance definition that reflects the true distance between two points or the natural length of a vector. In fields such as data analysis and machine learning, Euclidean distance is often used to calculate the distance between sample points, and then perform tasks such as clustering and classification.

[0024] Spatial interpolation is a geospatial data analysis technique that explores and analyzes the inherent laws of the collected sample point data, and then extrapolates these laws to the entire study area to convert them into surface data. The essence of spatial interpolation is to use the data information of known points to scientifically predict the data of unknown points through certain mathematical models and algorithms. This method has a wide range of applications in geographic information systems (GIS), environmental sciences, meteorology, geology and other fields. There are many methods of spatial interpolation, including nearest neighbor interpolation, linear interpolation, bilinear interpolation, spline interpolation, Kriging interpolation, etc. Different interpolation methods are suitable for different application scenarios and data characteristics. Choosing a suitable interpolation method can improve the accuracy and reliability of data prediction.

[0025] In this specification, a data processing method is provided. One or more embodiments of this specification also relate to a method for classifying urban features, a data processing device, an urban feature classification device, a computing device, a computer-readable storage medium, and a computer program product, which are described in detail one by one in the following embodiments.

[0026] In practical applications, existing hyperspectral regional data classification algorithms still face many challenges when dealing with urban land object monitoring tasks, such as the mixed pixel effect leading to reduced classification accuracy, and weak recognition of small-scale land objects such as temporary buildings and small green spaces. In addition, existing technologies lack efficient classification and change detection mechanisms that adapt to urban environments. Therefore, how to combine drone spectral data acquisition with deep learning algorithms to improve the accuracy of land object classification and dynamic monitoring has become the key to solving the technical problems of urban land object monitoring.

[0027] The data processing method provided in this embodiment, in order to improve the accuracy of ground object recognition and support efficient and accurate regional planning, can first collect hyperspectral regional data for the target area, and extract the target area features from the hyperspectral regional data; on this basis, in order to improve the efficiency of object classification, the target area features can be input into the object classification model to classify the regional objects in the target area, so as to obtain the object classification information of the regional objects in the target area according to the classification results; in order to enable the classification model to have a strong object boundary recognition ability, the recognition accuracy of the object classification model for the mixed pixel area can be optimized by the hierarchical sensitive spectral loss function in the training stage; so that the optimized object classification model can still have a strong boundary recognition accuracy for the data collected in the complex environment, so as to ensure that the object classification information of the regional objects in the target area output by the model is more accurate. Thereafter, in order to facilitate the use for regional planning and management, the hyperspectral regional data can be compared with the historical hyperspectral regional data of the target area to obtain the regional change information of the target area; and then the visual regional report corresponding to the target area can be constructed based on the object classification information and the regional change information. When the ground object analysis and processing is realized, the accuracy and efficiency can be effectively improved, so as to facilitate the use of downstream business for regional planning and management.

[0028] See also Figure 1 , Figure 1 A flow chart of a data processing method provided according to an embodiment of the present specification is shown, which specifically includes the following steps.

[0029] Step S102 : collecting hyperspectral region data for a target region, and extracting target region features from the hyperspectral region data.

[0030] The data processing method provided in this embodiment can be applied to object classification scenarios in any area, such as urban land object classification scenarios, housing item classification scenarios, park building classification scenarios, etc. This embodiment takes the urban land object classification scenario as an example to illustrate the data processing method. For descriptions of other scenarios, please refer to the same or corresponding descriptions in this embodiment, and this embodiment will not be repeated in detail.

[0031] Specifically, the target area specifically refers to the area that needs to be classified, such as the urban area; correspondingly, the hyperspectral area data specifically refers to the data obtained by setting the device to collect spectral data for the target area, which can be obtained by collecting the hyperspectral sensor configured by the drone, or by other devices configured with hyperspectral sensors, such as airplanes. Correspondingly, the target area feature specifically refers to the vector expression of the ground object information associated with the target area extracted from the hyperspectral area data.

[0032] Based on this, in order to improve the accuracy of object recognition and support efficient and accurate regional planning, we can first collect hyperspectral regional data for the target area and extract the target area features from the hyperspectral regional data; on this basis, in order to improve the efficiency of object classification, the target area features can be input into the object classification model to classify the regional objects in the target area, so as to obtain the object classification information of the regional objects in the target area according to the classification results; in order to enable the classification model to have a strong object boundary recognition ability, the object classification model can be optimized for the recognition accuracy of the mixed pixel area through the hierarchical sensitive spectral loss function during the training stage; so that the optimized object classification model can still have a strong boundary recognition accuracy for the data collected in the complex environment, so as to ensure that the object classification information of the regional objects in the target area output by the model is more accurate. After that, in order to facilitate the use of regional planning and management, the hyperspectral regional data can be compared with the historical hyperspectral regional data of the target area to obtain the regional change information of the target area; and then the visual regional report corresponding to the target area can be constructed based on the object classification information and the regional change information. When the object is analyzed and processed, the accuracy and efficiency can be effectively improved, so that it is convenient for downstream businesses to use it for regional planning and management.

[0033] Furthermore, in order to enable the drone to collect high-quality spectral data for use, the flight strategy of the drone can be planned in the following manner. In this embodiment, the specific implementation is as follows: A flight strategy is constructed for the UAV according to the spectral collection range information, spectral resolution information, spatial resolution information, and flight direction attribute information; and based on the flight strategy, the UAV is driven to collect hyperspectral area data for the target area.

[0034] Specifically, the collection range information refers to the range of multi-band spectral data collected by the hyperspectral sensor configured for the UAV for the target area, such as 400nm to 2500nm; correspondingly, the spectral resolution information refers to the resolution set by the hyperspectral sensor, such as 5nm; the spatial resolution refers to the resolution that can be achieved for collecting hyperspectral area data for the target area, such as 10cm; the flight attribute information refers to the flight control information corresponding to the UAV, including but not limited to flight speed, heading angle, altitude, etc.

[0035] Based on this, when using a drone to collect hyperspectral regional data for a target area, in order to ensure that the collected hyperspectral regional data carries more ground object information, a flight strategy can be constructed for the drone according to the spectral collection range information, spectral resolution information, spatial resolution information, and flight direction attribute information; thereafter, the drone can be driven to collect hyperspectral regional data for the target area based on the flight strategy, so that the hyperspectral regional data collected at this time can be used for subsequent ground object classification processing.

[0036] In summary, by combining multi-dimensional information to build a flight strategy for UAVs, when UAVs collect hyperspectral area data, the collected data can carry more ground object information, which is convenient for subsequent high-precision ground object classification.

[0037] Furthermore, in order to ensure the consistency of data collected in different time periods, the collected data needs to be preprocessed to ensure that the preprocessed data is highly similar. In this embodiment, the specific implementation method is as follows: Initial hyperspectral region data is collected for a target region by a hyperspectral sensor of an unmanned aerial vehicle; the initial hyperspectral region data is preprocessed to obtain hyperspectral region data, wherein the hyperspectral region data has a consistent relationship with the historical hyperspectral region data.

[0038] Specifically, preprocessing refers to operations such as data denoising and / or correction on the initial hyperspectral region data, so as to make the hyperspectral region data consistent with the historical hyperspectral region data collected in the historical period, so that it can be used for subsequent regional change information construction.

[0039] Based on this, in order to realize that the hyperspectral regional data can also be used for the regional change information of the target area after the classification and processing of the ground objects, the hyperspectral regional data of the target area can be collected by the hyperspectral sensor of the UAV, and then the initial hyperspectral regional data can be preprocessed, so that the hyperspectral regional data obtained after preprocessing has a consistent relationship with the historical hyperspectral regional data, which is convenient for subsequent use.

[0040] On this basis, illumination correction, denoising and difference correction processing can be used to implement preprocessing. In this embodiment, the specific implementation method is as follows: The initial hyperspectral region data is subjected to illumination correction processing to obtain intermediate hyperspectral region data; the intermediate hyperspectral region data is subjected to denoising processing using a Gaussian filter, and the denoised intermediate hyperspectral region data is subjected to viewing angle difference correction processing to obtain target hyperspectral region data; the signal-to-noise ratio of the target hyperspectral region data is calculated, and when it is determined according to the signal-to-noise ratio that the target hyperspectral region data meets a data quality condition, the target hyperspectral region data is used as the hyperspectral region data.

[0041] Specifically, illumination correction processing refers to the process of correcting the influence of external illumination on the initial hyperspectral region data, and the intermediate hyperspectral region data is the hyperspectral region data obtained after eliminating the influence of illumination. De-noising processing refers to the process of eliminating the noise interference carried in the intermediate hyperspectral region data, which is used to avoid the influence of noise, so that the processing result deviates from the actual result. Correspondingly, perspective difference correction processing refers to the operation of image alignment processing on the collected hyperspectral region data, which is mainly to correct the error caused by the inconsistent perspective of data collected from different perspectives. Correspondingly, data quality conditions refer to the conditions for detecting whether the pre-processed hyperspectral region data can be used for ground object classification.

[0042] Based on this, after obtaining the initial hyperspectral regional data collected by the UAV, in order to eliminate the influence of redundant information and make the processed hyperspectral regional data more able to reflect the ground object information in the region, first, the initial hyperspectral regional data can be subjected to illumination correction processing to eliminate the influence of illumination, and then obtain the intermediate hyperspectral regional data; secondly, the intermediate hyperspectral regional data can be denoised using a Gaussian filter to remove redundant information; the denoised intermediate hyperspectral regional data is subjected to perspective difference correction processing to correct the influence of perspective deviation and obtain the target hyperspectral regional data; finally, the signal-to-noise ratio of the target hyperspectral regional data can be calculated, and the data quality can be detected according to the signal-to-noise ratio, so that when it is determined that the target hyperspectral regional data meets the data quality conditions, the target hyperspectral regional data can be used as the hyperspectral regional data for subsequent ground object classification.

[0043] In practical applications, the collection and preprocessing of hyperspectral area data can use the hyperspectral sensor carried by drones to collect data in the target area. The collected data can be corrected for illumination, noise, and perspective differences to ensure that the data in different collection periods and regions are consistent and high-quality, providing a reliable basis for subsequent analysis and processing.

[0044] In specific implementation, for the urban land object classification scenario, the collection of hyperspectral area data can be carried out by using a drone equipped with a hyperspectral sensor to perform a coverage scan of the target urban area. The collection range can be set to multi-band spectral data from 400nm to 2500nm, with a spectral resolution of about 5nm and a spatial resolution of up to 10 cm. Among them, the sensors used cover the visible light to near-infrared bands, so as to achieve detailed recording of the land object information in the target urban area. At the same time, in order to ensure comprehensive data coverage, multiple flight paths with different headings and flight altitudes can be planned during the collection process. According to the area of ​​the target urban area and the sensor resolution, the flight altitude (such as 300 meters) and flight speed (such as 10m / s) of the drone are set to ensure that a sufficiently fine spatial resolution is obtained. The collected data is stored as a three-dimensional spectral data cube , that is, the hyperspectral region data, where is the spatial coordinate, is the wavelength.

[0045] Furthermore, after obtaining the hyperspectral region data, the collected hyperspectral region data needs to be preprocessed. In this process, illumination correction can be completed first to eliminate the influence of external illumination on the image. In practical applications, the FLAASH method can be used for atmospheric correction, which can be achieved through the following formula (1): (1) To adjust the spectral data, is the original data, is the blackbody radiation value, is the radiation conversion coefficient.

[0046] Furthermore, for noise removal, a Gaussian filter can be used to smooth the data to reduce noise interference. The filtering operation formula is as follows (2): (2) in, is the Gaussian kernel function, is the original image, is the filtered image, is the filter window size.

[0047] On this basis, the perspective difference correction can be performed; the images collected from different perspectives can be aligned through feature point matching and registration algorithms. During the registration process, the formula for minimizing the error is as follows (3): (3) in, and For two effects, is the coordinate of the matching point, is the total matching points.

[0048] Based on the above processing, the oblique images collected by the drone can be geometrically corrected, and the hyperspectral area data can be converted into a standard map coordinate system, and then an orthophoto can be generated. In specific implementation, the oblique images can be accurately mapped to the standard map coordinate system through projection transformation or orthophoto generation. Based on the attitude parameters and GPS positioning information of the drone, the geometric distortion of the image can be corrected to ensure that it meets the requirements of the standard map coordinate system. The geometric correction process can be implemented by the following formula (4): (4) in, is the original pixel coordinate, is the corrected coordinate, is the transformation parameter, calculated by the calibration model.

[0049] After completing the preprocessing of the hyperspectral region data, in order to ensure that the data can be used for subsequent calculations, the data quality can be evaluated to ensure the consistency and reliability of the data. The evaluation index is the signal-to-noise ratio (SNR), which measures the data quality by calculating the ratio of signal to noise. It can be achieved by the following formula (5): (5) in, is the signal mean, When it is determined through the above processing that the hyperspectral regional data collected for the target area can meet the subsequent ground object classification requirements, the subsequent processing operation can be performed.

[0050] In summary, by preprocessing the collected hyperspectral area data, the data can be guaranteed to have high quality and contain richer ground object information, thereby improving the accuracy of subsequent ground object classification.

[0051] After obtaining the hyperspectral region data, in order to ensure that the extracted features can fully reflect the relevant attributes of the target region, a convolutional neural network can be used to extract regional features. In this embodiment, the specific implementation method is as follows: The feature dimension of the hyperspectral region data is standardized to obtain model output information; the model input information is input into a compression-excited convolutional neural network for feature extraction to obtain target region features.

[0052] Specifically, the compression-excitation convolutional neural network refers to a convolutional neural network that performs feature processing on hyperspectral region data. Based on this, in order to ensure that the extracted target region features carry more accurate ground object information, the feature dimension of the hyperspectral region data can be standardized so that the model output information obtained after processing meets the model input requirements. Then the model input information can be input into the compression-excitation convolutional neural network for feature extraction, thereby obtaining the target region features for subsequent use.

[0053] In practical applications, by selecting a compressed excitation convolutional neural network to process hyperspectral area data, it can ensure that the model can efficiently capture the subtle differences between different land object categories, and by enhancing key features, the model's sensitivity to complex land object boundaries (such as the transition zone between buildings and green spaces) can be improved, thereby further optimizing the classification accuracy.

[0054] For example, a city plans to build a development zone. In order to carry out regional planning and management of the development zone, it is necessary to provide a high-precision urban area change report for each construction stage to facilitate subsequent work guidance. Therefore, in the report construction stage, it is necessary to first accurately classify the objects in the development zone, and then mark the changed area. In this process, the drone can be controlled to collect hyperspectral regional data for the development zone through the hyperspectral sensor carried according to the above scanning settings. After collecting the hyperspectral regional data, it can be pre-processed in combination with formulas (1) to (4) to eliminate the influence of light, noise interference and perspective correction, so as to obtain higher quality hyperspectral regional data; further, after obtaining the hyperspectral regional data, the corresponding signal-to-noise ratio S can be calculated by formula (5), and compared with the set threshold. If it is greater than the threshold, it means that the quality of the pre-processed hyperspectral regional data is high, so it can be used for subsequent classification of objects and construction of regional change reports for the development zone.

[0055] In summary, by using the compressed excitation convolutional neural network to extract target area features, the model can better capture the differences between objects in the area, thereby improving the subsequent object classification accuracy.

[0056] Step S104, inputting the target area features into an object classification model to classify the area objects in the target area, and obtaining object classification information of the area objects in the target area, wherein the object classification model optimizes the recognition accuracy of the object classification model for mixed pixel areas through a hierarchical sensitive spectral loss function.

[0057] Specifically, after obtaining the hyperspectral regional data corresponding to the target area as described above, in order to accurately identify the object classification information of the regional objects in the target area, an object classification model can be used to determine the object classification information. Specifically, the target area features can be input into the object classification model to classify the regional objects in the target area, so that the object classification information of the regional objects in the target area can be output through the classification model. In order to improve the classification accuracy of regional objects through the object classification model, a hierarchical sensitive spectral loss function can be used to optimize the recognition accuracy of the object classification model for mixed pixel areas, so that the model can have a strong recognition ability for the boundary position of regional objects in the prediction stage, so that each regional object can be accurately assigned, such as lakes, traffic lights, parks, bridges, etc., for subsequent use.

[0058] Among them, the object classification model specifically refers to a model that inputs the target area features corresponding to the target area and outputs the object classification information corresponding to the regional object. The model can be optimized by the hierarchical sensitive spectral loss function so that the model can learn the ability to divide the boundaries of different regional objects. The hierarchical sensitive spectral loss function specifically refers to the loss function used by the model in the training stage. This function can enhance the model's ability to recognize the junctions of different regional objects, thereby distinguishing different regional objects and outputting the real object classification information corresponding to each regional object. Correspondingly, the mixed pixel area is the area where different regional objects meet.

[0059] Furthermore, when classifying the regional objects in the target area using the object classification model, in order to improve the classification accuracy, multi-scale and multi-level processing can be used. In this embodiment, the specific implementation method is as follows: The target area feature is input into the object classification model, and the hyperspectral cube feature corresponding to the regional object in the target area feature is convolved by the multi-scale convolution kernel in the object classification model to obtain a feature map corresponding to the hyperspectral cube feature; the feature map is dynamically reweighted by the compression excitation unit in the object classification model, and the multi-scale feature and multi-level feature corresponding to the hyperspectral cube feature are determined according to the processing result; the multi-scale feature and the multi-level feature are spliced ​​into features to be classified, and the classification unit in the object classification model is used to process the features to be classified to obtain object classification information of the regional object in the target area.

[0060] Specifically, multi-scale convolution kernel refers to convolution kernels corresponding to multiple different scales; accordingly, hyperspectral cube features refer to feature expressions corresponding to different pixels in the target area feature, and the feature carries the band information corresponding to the pixel. Accordingly, feature map refers to the feature map obtained after convolution processing. Accordingly, dynamic reweighting refers to the operation of recalculating weights and weighted processing, which is achieved through compression excitation units. Accordingly, multi-scale features and multi-level features refer to feature expressions corresponding to the height and depth of the target area.

[0061] Based on this, when classifying regional objects in the target area through the object classification model, the target area features can be first input into the object classification model, and the hyperspectral cube features corresponding to the regional objects in the target area features can be convolved through the multi-scale convolution kernel in the object classification model, and the feature map corresponding to the hyperspectral cube features can be obtained according to the processing results; thereafter, the feature map can be dynamically re-weighted through the compression excitation unit in the object classification model, so that the multi-scale features and multi-level features corresponding to the hyperspectral cube features can be determined according to the processing results; on this basis, the multi-scale features and multi-level features can be spliced ​​into features to be classified, so as to realize the processing of the features to be classified by the classification unit in the object classification model, and the object classification information of the regional objects in the target area can be obtained for subsequent use.

[0062] In practical applications, when performing multi-scale classification and fine-grained object recognition, various types of objects in urban environments can be classified through multi-scale convolutional networks (object classification models). This process can optimize objects with large scale differences such as high-rise buildings, roads and green spaces in cities, thereby ensuring that small-scale objects (such as temporary buildings and small green spaces) can also be accurately identified and improve the overall classification effect.

[0063] In the specific implementation, for the urban land object classification scenario, the feature dimension of the pre-processed hyperspectral data can be standardized to ensure that the input features are within the same scale range, which helps to accelerate model convergence and improve training stability. In the specific implementation, the Z-score standardization method can be used, and the standardization formula is as follows (6): (6) in, and They are The mean and standard deviation of the band. The standardized data can effectively reduce the numerical differences between different bands and ensure that the convolution operation can learn the characteristics of all bands in a balanced manner.

[0064] On this basis, the convolutional neural network can be used to perform preliminary feature extraction on hyperspectral regional data, so as to effectively capture the subtle differences between different types of objects and improve the sensitivity to complex object boundaries by enhancing key features. Especially in urban environments, it helps to distinguish between buildings, green spaces and other objects, and provides valuable features for subsequent classification tasks.

[0065] In this process, in order to further improve the classification accuracy and optimize the recognition of objects of different scales in urban environments, a multi-scale convolutional network can be used. By using convolution kernels of different sizes, the model can capture the features of objects at multiple scales, especially for small-scale objects and fine-grained areas with significant scale differences, which improves the recognition ability. Each layer of convolution kernel extracts corresponding features according to its scale characteristics to ensure that details from macro to micro are effectively captured. Specifically, the input data is a preprocessed hyperspectral cube. , the convolution operation uses multi-scale convolution kernels (e.g. and ), which aims to capture the features of objects at different scales. The convolution calculation formula is as follows (7): (7) in, is the output feature map, For the convolution kernels, is the bias, is the activation function (such as ReLU).

[0066] After that, the compression excitation module is introduced to dynamically reweight the features between channels to highlight the information of key bands. Specifically, the weight of each channel can be calculated by global average pooling, which can be achieved by the following formula (8): (8) in, For Channel The global feature description of and are the height and width of the feature map respectively. Further, two fully connected layers and activation functions are used to generate the channel weight vector , and weight the original features, which is processed by the following formula (9): (9) Among them, i and j are the spatial coordinate parameters of the feature map, i represents the vertical coordinate of the feature map, j represents the horizontal coordinate of the feature map, and i and j together define the position of each pixel in the feature map. Therefore, multiple convolutional layers can be used to extract deep features from weighted features, thereby further improving the network's ability to distinguish complex terrain areas. In addition, in order to avoid overfitting, regularization measures can be introduced after the convolution operation.

[0067] Finally, the extracted multi-scale and multi-level features are integrated. That is, the feature maps output by all convolutional layers are connected in the channel dimension to form a global feature vector, which is as follows (10): (10) in, represents the feature concatenation operation, For the The integrated feature vector will be used as the input of the subsequent classification module to provide the model with richer spectral and spatial information. On this basis, each feature vector is processed by the classification layer in the model to obtain the classification information corresponding to each feature in the urban area, so as to facilitate the subsequent construction of visualization reports.

[0068] In summary, by using the object classification model to process the target area features corresponding to the target area, it is possible to enhance the recognition accuracy of objects of different scales from multiple angles during the processing process, thereby ensuring that the final output object classification information is more in line with the actual situation.

[0069] In addition, in order to enable the object classification model to have a higher recognition accuracy and to be able to clear the object boundaries in the region for analysis, the two loss functions can be combined for model optimization. In this embodiment, the training of the object classification model includes: Obtain object classification samples, and input the object classification samples into the initial object classification model for processing to obtain predicted classification information; calculate a boundary loss value based on the predicted classification information and the sample classification information corresponding to the object classification samples according to a boundary recognition loss function, and calculate an object loss value based on the predicted classification information and the sample classification information according to an object recognition loss function; optimize the initial object classification model based on the boundary loss value and the object loss value until the object classification model that meets the training stop condition is obtained; wherein the boundary recognition loss function and the object recognition loss function constitute the hierarchical sensitive spectrum loss function, and the recognition accuracy of the initial object classification model for object boundaries is optimized by the boundary recognition loss function, and the recognition accuracy of the initial object classification model for regional objects is optimized by the object recognition loss function.

[0070] Specifically, the object classification sample specifically refers to the sample that has the same characteristic structure as the target area corresponding to the target area but is used as the model training stage. Correspondingly, the predicted classification information specifically refers to the classification information obtained after identifying the sample area object in the sample. The boundary recognition loss function specifically refers to the loss function that strengthens the accuracy of model learning object boundary recognition. The object recognition loss function specifically refers to the loss function of the object recognition accuracy of the reinforcement model learning. Correspondingly, the training stop condition specifically refers to the condition for stopping the training of the object classification model, which includes but is not limited to the loss value comparison condition, the number of iterations condition or the verification set verification condition, etc.; in the specific implementation, it can be selected according to actual needs, and this embodiment does not make any limitation here.

[0071] Based on this, in order to enable the object classification model to have a strong object recognition ability in the application stage, it can not only accurately identify each regional object, but also effectively divide different regional objects for subsequent construction of visualization reports, and can be trained using a dual loss function. Specifically, the object classification sample can be obtained first, and input into the initial object classification model for processing to obtain the predicted classification information; after obtaining the predicted classification information, according to the label corresponding to the sample, the boundary loss value can be calculated based on the predicted classification information and the sample classification information corresponding to the object classification sample according to the boundary recognition loss function, and the object loss value can be calculated based on the predicted classification information and the sample classification information according to the object recognition loss function; the boundary loss value can reflect the accuracy of the object classification model in the current stage for boundary recognition, and the object loss value can reflect the accuracy of the object classification model in the current stage for object recognition, so the initial object classification model can be optimized based on the boundary loss value and the object loss value; if the optimized model does not meet the training stop condition, new samples can be selected to train it until the object classification model that meets the training stop condition is obtained and deployed for use. Among them, the boundary recognition loss function and the object recognition loss function constitute a hierarchical sensitive spectral loss function, and the recognition accuracy of the initial object classification model for object boundaries is optimized by the boundary recognition loss function, and the recognition accuracy of the initial object classification model for regional objects is optimized by the object recognition loss function.

[0072] In practical applications, when optimizing object classification models, in order to improve the recognition accuracy of the model at the boundary, the hierarchical sensitive spectral loss function can be used to optimize the classification accuracy of hyperspectral region data, especially in complex mixed pixel areas. By improving the recognition ability of the boundaries of different types of objects such as buildings and green spaces, the problem of confusion of subdivision categories in hyperspectral region data can be solved to ensure the accuracy of classification results.

[0073] In specific implementation, for urban land object classification scenarios, the model training can first optimize the classification accuracy for the mixed pixel area in the hyperspectral area data. By weighting each layer of features, the model's ability to recognize the boundaries of land objects such as buildings and green spaces is improved. The specific loss function formula is as follows (11): (11) in, For the The loss of layers, is the weighting coefficient.

[0074] Furthermore, for mixed pixel areas, the local sensitive spectral loss function can be used Strengthen the recognition of fine-grained objects. The loss function formula is as follows (12): (12) in, To predict the spectrum, is the real spectrum, is the category weighting coefficient.

[0075] On this basis, the classification loss and the optimized hierarchical sensitive spectrum loss can be combined to obtain the final optimized loss function, as shown in the following formula (13): (13) The optimized model can effectively improve the classification accuracy of complex mixed pixel areas and enhance the overall classification effect of hyperspectral data.

[0076] Continuing with the above example, after obtaining the high-quality hyperspectral regional data corresponding to the development zone, the feature expression V corresponding to the hyperspectral regional data can be extracted first, and then it can be input into the feature classification model trained and optimized by formulas (11) to (13) for processing. After inputting it into the feature classification model, the model can complete the classification processing of each feature in the development zone in combination with the above formulas (6) to (10). After the classification processing, it can be realized on the development zone map matching the hyperspectral regional data, according to the mapping relationship between the two, the type information of the feature corresponding to each position can be marked for subsequent use.

[0077] In summary, by using boundary recognition loss function and object recognition loss function to optimize the object classification model, the model can improve the recognition accuracy of object boundaries while ensuring the object recognition accuracy, thereby effectively improving the accuracy of object classification in any scenario in the application stage.

[0078] Step S106: Compare the hyperspectral region data with the historical hyperspectral region data of the target region to obtain region change information of the target region.

[0079] Specifically, after obtaining the object classification information corresponding to each regional object in the target area, it means that the object classification processing for the regional objects in the target area has been completed. In order to provide accurate reference for the planning and management of the target area, the hyperspectral regional data can be compared with the historical hyperspectral regional data of the target area, and the regional change information of the target area can be obtained according to the comparison results. Subsequently, the visual regional report of the corresponding target area is constructed by combining the regional change information and the object classification information, which can provide a reference for the planning and management of the target area.

[0080] Among them, the historical hyperspectral regional data specifically refers to the hyperspectral regional data collected for the target area in the previous acquisition cycle according to the same settings as the hyperspectral regional data. By comparing the hyperspectral regional data collected in the current acquisition cycle with the historical hyperspectral regional data, the regional changes in the target area between the two acquisition cycles can be determined for subsequent report construction. Among them, the regional change information specifically refers to the information on regional changes reflected in the visualization dimension obtained by comparing the hyperspectral regional data of two adjacent cycles, which can be understood as the image difference information after spectral image comparison.

[0081] Furthermore, after obtaining the regional change information, in order to avoid the impact caused by inaccurate data collection, regional change condition detection can be performed before constructing the visual regional report, thereby improving accuracy. In this embodiment, the specific implementation method is as follows: Determine at least two change areas corresponding to the area change information, and perform spatial aggregation on the at least two change areas to obtain a global change area; detect whether the global change area meets the area change condition corresponding to the target area; if so, execute the step of constructing a visual area report corresponding to the target area based on the object classification information and the area change information.

[0082] Specifically, the global change region refers to a change region obtained by spatially aggregating at least two change regions corresponding to the region change information. Correspondingly, the region change condition refers to a condition for detecting whether the change region corresponding to the region change information determined in the current period meets the actual situation.

[0083] Based on this, when constructing a visual regional report, at least two change regions corresponding to the regional change information can be determined first. At this time, spatial aggregation can be performed on at least two change regions to obtain a global change region; then, it is detected whether the global change region meets the regional change condition corresponding to the target region; if not, it means that the regional change information determined by the data collected in the current period through processing does not conform to the actual situation, which further indicates that the data may be wrong, so data collection and processing can be performed again. If so, step S108 can be executed.

[0084] In practical applications, time series analysis technology can be used to compare hyperspectral regional data at different time nodes to identify changes in urban areas. Through the change detection module, new buildings, green space occupation and other change areas can be accurately located, and detailed change reports can be generated to provide decision-making basis for dynamic urban management.

[0085] That is to say, for the urban land object classification scenario, time series analysis technology can be used to compare and analyze the hyperspectral area data at different time nodes. By calculating the change of each pixel in the time dimension, the area with significant changes can be identified. The specific change detection method can be expressed as the following formula (14): (14) in, and The time nodes and Hyperspectral data, For the band, Indicates pixel position Spectral differences at different time points.

[0086] Furthermore, the change detection module can be used to accurately locate the changed area. , the difference image is binarized to identify the changed areas such as new buildings and green space occupation. The processing can be achieved by the following formula (15): (15) If the difference Exceeds a preset threshold , then it is considered that the position has changed. The threshold It is obtained through the analysis of historical data and is used to distinguish between changed areas and unchanged areas.

[0087] Furthermore, spatial aggregation of all the changed areas can be used to further analyze whether the changed areas are consistent with the actual changes, such as new buildings, green space expansion, etc. At this time, the change value of each pixel Spatial aggregation is performed to obtain the overall change value of each region, which is calculated using the following formula (16): (16) if Greater than a global threshold , then it is considered that the area has changed. After obtaining the regional change information, a visual regional report can be constructed for the target area for use in planning and managing the target area.

[0088] Using the above example, after obtaining the classification information corresponding to each land object in the development zone, in order to determine the construction changes in the development zone between the current cycle and the previous cycle, the historical hyperspectral regional data corresponding to the development zone can be obtained. Then, the regional change information corresponding to the development zone in two adjacent cycles can be calculated by the above formulas (14) to (16), so that the classification information of the land object can be combined to generate a change report of the development zone for the planning and management of the development zone.

[0089] In summary, by comparing hyperspectral data of adjacent periods to determine regional change information, the corresponding changes in the target area can be accurately determined for use in building reports.

[0090] Step S108: constructing a visualized area report corresponding to the target area based on the object classification information and the area change information.

[0091] Specifically, after obtaining the regional change information and the object classification information corresponding to the regional object, in order to provide a reference for the planning and management of the target area, a visual regional report corresponding to the target area can be constructed by combining the object classification information and the regional change information. The visual regional report specifically refers to a report that intuitively reflects the regional change situation of the target area and the corresponding type of each regional object, so as to be used for the planning and management of the target area.

[0092] Furthermore, when constructing a visual regional report, the fusion of classification information and change information can be achieved by constructing and updating a classification map. In this embodiment, the specific implementation method is as follows: A classification map corresponding to the target area is constructed according to the object classification information; the classification map is updated according to the area change information to obtain an initial change area marking map; the initial change area marking map is smoothed using a preset spatial difference algorithm, and a visualization area report corresponding to the target area is constructed according to the smoothing result.

[0093] Specifically, the classification map refers to a map constructed for the target area based on the object classification information, and the map carries the type description of each area object. Correspondingly, the initial change area marking map refers to a map that records the changes in the area. Correspondingly, smoothing processing refers to repairing the initial change area marking map so that the visual display of the map better meets the viewing requirements.

[0094] Based on this, when constructing a visual area report corresponding to the target area, you can first construct a classification map corresponding to the target area based on the object classification information; then update the classification map according to the area change information to obtain the initial change area marking map, and then realize carrying the area change information and object classification information in the map. Finally, you can use the preset spatial difference algorithm to smooth the initial change area marking map to realize the construction of a visual area report corresponding to the target area based on the smoothing result.

[0095] In practical applications, when generating a visual regional report, the changed area and the classification information corresponding to each regional object can be marked in the spatial distribution map of the target area, so as to facilitate the intuitive display of classification combinations and monitoring areas for regional management and planning.

[0096] In the specific implementation, for the urban feature classification scenario, a classification map can be constructed based on the above classification results and change detection results. ,in Indicates location The classification basis can be based on the distance between each category in the feature space using the Euclidean distance metric. Specifically, for each pixel , calculate the Euclidean distance between its spectral feature vector and the center vector of each feature category, which can be calculated using the following formula (17): (17) in, is the location in the hyperspectral region data Place The spectral value of the band, Yes Category exist The mean value on the band, is the number of spectral bands. By calculating the Euclidean distance between each pixel and the center of each category, the pixel will be classified into the category with the smallest distance.

[0097] Furthermore, the change detection results can be combined with the classification results to generate a change area marking map. By marking the change areas in the classification results, the change areas such as new buildings and green space expansion can be effectively identified and marked. The calculation formula (18) is as follows: (18) in, is a marker for the changed area, 1 means that the area has changed, Indicates the original category label.

[0098] Based on this, the spatial interpolation method can be used to smooth the classification results to avoid noise or unclear boundaries that may exist in the classification process. For example, the bilinear interpolation method can be used to smooth the classification results to make the spatial distribution map more coherent. The interpolation formula (19) is as follows: (19) in, is the classification result after smoothing. is the position of the neighboring pixels.

[0099] On this basis, a visualization report can be generated. In specific implementation, the classification results and the changed areas can be displayed intuitively through color coding and symbol annotation. For example, different types of objects are marked with different colors, and the changed areas can be highlighted for easy observation. The generation formula (20) of the visualization map is as follows: (20) in, For the final visualization image, The function is used to combine the classification results and the change area for visualization.

[0100] The resulting visual regional report will show the spatial distribution of various urban features and mark areas where changes have occurred, helping urban management departments to intuitively understand urban development and changes, and further support urban planning and decision-making.

[0101] Using the above example, after determining the regional change information and feature classification information, the above formulas (17) to (20) can be combined to complete the construction of a visual regional report on the development zone. The report can record the areas that have changed in the development zone and the description of the areas that have been completed, thereby facilitating the management department to manage and plan the development zone.

[0102] The data processing method provided in this embodiment, in order to improve the accuracy of ground object recognition and support efficient and accurate regional planning, can first collect hyperspectral regional data for the target area, and extract the target area features from the hyperspectral regional data; on this basis, in order to improve the efficiency of object classification, the target area features can be input into the object classification model to classify the regional objects in the target area, so as to obtain the object classification information of the regional objects in the target area according to the classification results; in order to enable the classification model to have a strong object boundary recognition ability, the recognition accuracy of the object classification model for the mixed pixel area can be optimized by the hierarchical sensitive spectral loss function in the training stage; so that the optimized object classification model can still have a strong boundary recognition accuracy for the data collected in the complex environment, so as to ensure that the object classification information of the regional objects in the target area output by the model is more accurate. Thereafter, in order to facilitate the use for regional planning and management, the hyperspectral regional data can be compared with the historical hyperspectral regional data of the target area to obtain the regional change information of the target area; and then the visual regional report corresponding to the target area can be constructed based on the object classification information and the regional change information. When the ground object analysis and processing is realized, the accuracy and efficiency can be effectively improved, so as to facilitate the use of downstream business for regional planning and management.

[0103] See also Figure 2 , Figure 2 A flow chart of a method for classifying urban features provided according to an embodiment of the present specification is shown, which specifically includes the following steps.

[0104] Step S202 : collecting hyperspectral region data for urban areas, and extracting urban region features from the hyperspectral region data.

[0105] Step S204, inputting the urban area features into a classification model to classify the ground objects in the urban area, and obtaining ground object classification information of the ground objects in the urban area, wherein the classification model optimizes the recognition accuracy of the classification model for mixed pixel areas through a hierarchical sensitive spectral loss function.

[0106] Step S206: Compare the hyperspectral region data with the historical hyperspectral region data of the urban area to obtain regional change information of the urban area.

[0107] Step S208: constructing an urban area change report corresponding to the urban area based on the land feature classification information and the area change information.

[0108] The urban feature classification method provided in this embodiment is applied to the classification scenario of features in urban areas, so that the classification results can be used for city planning and management. The description of the urban feature classification method can refer to the same or corresponding description content in the above embodiments, and this embodiment will not be elaborated in detail here.

[0109] The following combination Figure 3 , taking the application of the data processing method provided in this specification in the urban planning scenario as an example, the data processing method is further explained. Figure 3 A processing flow chart of a data processing method provided by an embodiment of the present specification is shown, which specifically includes the following steps.

[0110] Step S302: collect initial hyperspectral region data for the target area through the hyperspectral sensor of the drone, perform illumination correction processing on the initial hyperspectral region data, and obtain intermediate hyperspectral region data.

[0111] Step S304: denoising the intermediate hyperspectral region data using a Gaussian filter, and performing a viewing angle difference correction process on the denoised intermediate hyperspectral region data to obtain target hyperspectral region data.

[0112] Step S306, calculating the signal-to-noise ratio of the target hyperspectral region data, and determining that the target hyperspectral region data meets the data quality condition according to the signal-to-noise ratio, taking the target hyperspectral region data as the hyperspectral region data, wherein the hyperspectral region data has a consistent relationship with the historical hyperspectral region data.

[0113] Step S308, perform standardization processing on the feature dimension of the hyperspectral region data to obtain model output information, input the model input information into the compression-excitation convolutional neural network for feature extraction, and obtain the target region features.

[0114] Step S310, inputting the target region feature into the object classification model, performing convolution processing on the hyperspectral cube feature of the corresponding region object in the target region feature through the multi-scale convolution kernel in the object classification model, and obtaining a feature map corresponding to the hyperspectral cube feature.

[0115] Step S312, dynamically reweighting the feature map through the compression excitation unit in the object classification model, and determining the multi-scale features and multi-level features corresponding to the hyperspectral cube features according to the processing results.

[0116] Step S314, concatenating the multi-scale features and the multi-level features into features to be classified, and using the classification unit in the object classification model to process the features to be classified, to obtain object classification information of the regional objects in the target area.

[0117] Step S316: compare the hyperspectral region data with the historical hyperspectral region data of the target region to obtain the region change information of the target region.

[0118] Step S318, constructing a classification map corresponding to the target area according to the object classification information, updating the classification map according to the area change information, and obtaining an initial change area marking map.

[0119] Step S320, using a preset spatial difference algorithm to smooth the initial change area marking map, and constructing a visual area report corresponding to the target area according to the smoothing result.

[0120] In summary, in order to improve the accuracy of object recognition and support efficient and accurate regional planning, we can first collect hyperspectral regional data for the target area and extract the target area features from the hyperspectral regional data; on this basis, in order to improve the efficiency of object classification, the target area features can be input into the object classification model to classify the regional objects in the target area, so as to obtain the object classification information of the regional objects in the target area according to the classification results; in order to enable the classification model to have a strong object boundary recognition ability, the object classification model can be optimized for the recognition accuracy of the mixed pixel area through the hierarchical sensitive spectral loss function during the training stage; so that the optimized object classification model can still have a strong boundary recognition accuracy for the data collected in the complex environment, so as to ensure that the object classification information of the regional objects in the target area output by the model is more accurate. After that, in order to facilitate the use of regional planning and management, the hyperspectral regional data can be compared with the historical hyperspectral regional data of the target area to obtain the regional change information of the target area; and then the visual regional report corresponding to the target area can be constructed based on the object classification information and regional change information. When the object is analyzed and processed, the accuracy and efficiency can be effectively improved, so that it is convenient for downstream businesses to use it for regional planning and management.

[0121] Corresponding to the above method embodiment, this specification also provides a data processing device embodiment, Figure 4 FIG. 1 is a schematic diagram showing the structure of a data processing device provided by an embodiment of the present specification. Figure 4 As shown, the device comprises: The acquisition module 402 is configured to acquire hyperspectral region data for a target region and extract target region features from the hyperspectral region data; An input module 404 is configured to input the target area feature into an object classification model to classify the area object in the target area, and obtain object classification information of the area object in the target area, wherein the object classification model optimizes the recognition accuracy of the object classification model for the mixed pixel area through a hierarchical sensitive spectral loss function; A comparison module 406 is configured to compare the hyperspectral region data with historical hyperspectral region data of the target region to obtain region change information of the target region; The construction module 408 is configured to construct a visual area report corresponding to the target area based on the object classification information and the area change information.

[0122] In an optional embodiment, the acquisition module 402 is further configured to: Initial hyperspectral region data is collected for a target region by a hyperspectral sensor of an unmanned aerial vehicle; the initial hyperspectral region data is preprocessed to obtain hyperspectral region data, wherein the hyperspectral region data has a consistent relationship with the historical hyperspectral region data.

[0123] In an optional embodiment, the acquisition module 402 is further configured to: The initial hyperspectral region data is subjected to illumination correction processing to obtain intermediate hyperspectral region data; the intermediate hyperspectral region data is subjected to denoising processing using a Gaussian filter, and the denoised intermediate hyperspectral region data is subjected to viewing angle difference correction processing to obtain target hyperspectral region data; the signal-to-noise ratio of the target hyperspectral region data is calculated, and when it is determined according to the signal-to-noise ratio that the target hyperspectral region data meets a data quality condition, the target hyperspectral region data is used as the hyperspectral region data.

[0124] In an optional embodiment, the acquisition module 402 is further configured to: The feature dimension of the hyperspectral region data is standardized to obtain model output information; the model input information is input into a compression-excited convolutional neural network for feature extraction to obtain target region features.

[0125] In an optional embodiment, the input module 404 is further configured to: The target area feature is input into the object classification model, and the hyperspectral cube feature corresponding to the regional object in the target area feature is convolved by the multi-scale convolution kernel in the object classification model to obtain a feature map corresponding to the hyperspectral cube feature; the feature map is dynamically reweighted by the compression excitation unit in the object classification model, and the multi-scale feature and multi-level feature corresponding to the hyperspectral cube feature are determined according to the processing result; the multi-scale feature and the multi-level feature are spliced ​​into features to be classified, and the classification unit in the object classification model is used to process the features to be classified to obtain object classification information of the regional object in the target area.

[0126] In an optional embodiment, the training of the object classification model includes: Obtain object classification samples, and input the object classification samples into the initial object classification model for processing to obtain predicted classification information; calculate a boundary loss value based on the predicted classification information and the sample classification information corresponding to the object classification samples according to a boundary recognition loss function, and calculate an object loss value based on the predicted classification information and the sample classification information according to an object recognition loss function; optimize the initial object classification model based on the boundary loss value and the object loss value until the object classification model that meets the training stop condition is obtained; wherein the boundary recognition loss function and the object recognition loss function constitute the hierarchical sensitive spectrum loss function, and the recognition accuracy of the initial object classification model for object boundaries is optimized by the boundary recognition loss function, and the recognition accuracy of the initial object classification model for regional objects is optimized by the object recognition loss function.

[0127] In an optional embodiment, the device further comprises: The detection module is configured to determine at least two change areas corresponding to the area change information, and perform spatial aggregation on the at least two change areas to obtain a global change area; detect whether the global change area meets the area change condition corresponding to the target area; if so, execute the step of constructing a visual area report corresponding to the target area based on the object classification information and the area change information.

[0128] In an optional embodiment, the construction module 408 is further configured to: A classification map corresponding to the target area is constructed according to the object classification information; the classification map is updated according to the area change information to obtain an initial change area marking map; the initial change area marking map is smoothed using a preset spatial difference algorithm, and a visualization area report corresponding to the target area is constructed according to the smoothing result.

[0129] In an optional embodiment, the acquisition module 402 is further configured to: A flight strategy is constructed for the UAV according to the spectral collection range information, spectral resolution information, spatial resolution information, and flight direction attribute information; and based on the flight strategy, the UAV is driven to collect hyperspectral area data for the target area.

[0130] The data processing device provided in this embodiment can first collect hyperspectral regional data for the target area in order to improve the accuracy of object recognition and support efficient and accurate regional planning, and extract the target area features from the hyperspectral regional data; on this basis, in order to improve the efficiency of object classification, the target area features can be input into the object classification model to classify the regional objects in the target area, so as to obtain the object classification information of the regional objects in the target area according to the classification results; in order to enable the classification model to have a strong object boundary recognition ability, the recognition accuracy of the object classification model for the mixed pixel area can be optimized by the hierarchical sensitive spectral loss function in the training stage; so that the optimized object classification model can still have a strong boundary recognition accuracy for the data collected in the complex environment, so as to ensure that the object classification information of the regional objects in the target area output by the model is more accurate. Thereafter, in order to facilitate the use for regional planning and management, the hyperspectral regional data can be compared with the historical hyperspectral regional data of the target area to obtain the regional change information of the target area; and then the visual regional report corresponding to the target area can be constructed based on the object classification information and the regional change information. When the object analysis and processing is realized, the accuracy and efficiency can be effectively improved, so as to facilitate the use of downstream business for regional planning and management.

[0131] The above is a schematic scheme of a data processing device of this embodiment. It should be noted that the technical scheme of the data processing device and the technical scheme of the above data processing method belong to the same concept, and the details of the technical scheme of the data processing device that are not described in detail can be referred to the description of the technical scheme of the above data processing method.

[0132] Corresponding to the above method embodiment, this specification also provides an urban feature classification device embodiment, Figure 5 FIG. 2 shows a schematic diagram of the structure of a device for classifying urban features provided by an embodiment of the present specification. Figure 5 As shown, the device comprises: The data collection module 502 is configured to collect hyperspectral region data for urban areas and extract urban region features from the hyperspectral region data; The input model module 504 is configured to input the urban area features into the classification model to classify the ground objects in the urban area, and obtain ground object classification information of the ground objects in the urban area, wherein the classification model optimizes the recognition accuracy of the classification model for the mixed pixel area through a hierarchical sensitive spectral loss function; The data comparison module 506 is configured to compare the hyperspectral region data with the historical hyperspectral region data of the urban region to obtain the regional change information of the urban region; The report building module 508 is configured to build an urban area change report corresponding to the urban area based on the feature classification information and the area change information.

[0133] The above is a schematic scheme of a city feature classification device of this embodiment. It should be noted that the technical scheme of the city feature classification device and the technical scheme of the above-mentioned city feature classification method belong to the same concept, and the details not described in detail in the technical scheme of the city feature classification device can be referred to the description of the technical scheme of the above-mentioned city feature classification method.

[0134] Figure 6 The block diagram of a computing device 600 according to an embodiment of the present specification is shown. The components of the computing device 600 include but are not limited to a memory 610 and a processor 620. The processor 620 is connected to the memory 610 via a bus 630, and the database 650 is used to store data.

[0135] The computing device 600 also includes an access device 640 that enables the computing device 600 to communicate via one or more networks 660. Examples of these networks include a public switched telephone network (PSTN), a local area network (LAN), a wide area network (WAN), a personal area network (PAN), or a combination of communication networks such as the Internet. The access device 640 may include one or more of any type of network interface (e.g., a network interface card (NIC)) that is wired or wireless, such as an IEEE 802.11 wireless local area network (WLAN) wireless interface, a world-wide interoperability for microwave access (Wi-MAX) interface, an Ethernet interface, a universal serial bus (USB) interface, a cellular network interface, a Bluetooth interface, and a near field communication (NFC).

[0136] In one embodiment of the present specification, the above components of the computing device 600 and Figure 6 Other components not shown in the figure may also be connected to each other, for example, via a bus. It should be understood that Figure 6The computing device structure block diagram shown is only for the purpose of illustration, and is not intended to limit the scope of this specification. Those skilled in the art can add or replace other components as needed.

[0137] The computing device 600 may be any type of stationary or mobile computing device, including a mobile computer or mobile computing device (e.g., a tablet computer, a personal digital assistant, a laptop computer, a notebook computer, a netbook, etc.), a mobile phone (e.g., a smart phone), a wearable computing device (e.g., a smart watch, smart glasses, etc.), or other types of mobile devices, or a stationary computing device such as a desktop computer or a personal computer (PC). The computing device 600 may also be a mobile or stationary server.

[0138] The processor 620 is used to execute the following computer executable instructions, which, when executed by the processor, implement the steps of the above-mentioned data processing method or urban feature classification method.

[0139] The above is a schematic scheme of a computing device of this embodiment. It should be noted that the technical scheme of the computing device and the technical scheme of the above-mentioned data processing method or urban feature classification method belong to the same concept, and the details not described in detail in the technical scheme of the computing device can be referred to the description of the technical scheme of the above-mentioned data processing method or urban feature classification method.

[0140] An embodiment of the present specification further provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the above-mentioned data processing method or urban feature classification method.

[0141] The above is a schematic scheme of a computer-readable storage medium of this embodiment. It should be noted that the technical scheme of the storage medium and the technical scheme of the above-mentioned data processing method or urban feature classification method belong to the same concept, and the details not described in detail in the technical scheme of the storage medium can be referred to the description of the technical scheme of the above-mentioned data processing method or urban feature classification method.

[0142] An embodiment of the present specification also provides a computer program product, including a computer program or instructions, which implement the steps of the above-mentioned data processing method or urban feature classification method when executed by a processor.

[0143] The above is a schematic scheme of a computer program product of this embodiment. It should be noted that the technical scheme of the computer program product and the technical scheme of the above-mentioned data processing method or urban feature classification method belong to the same concept, and the details not described in detail in the technical scheme of the computer program product can be referred to the description of the technical scheme of the above-mentioned data processing method or urban feature classification method.

[0144] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0145] The computer instructions include computer program codes, which may be in source code form, object code form, executable files or some intermediate forms, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium may be appropriately increased or decreased according to the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0146] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the embodiments of this specification are not limited by the order of the actions described, because according to the embodiments of this specification, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the embodiments of this specification.

[0147] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0148] The preferred embodiments of this specification disclosed above are only used to help explain this specification. The optional embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of the embodiments of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the embodiments of this specification, so that technicians in the relevant technical field can understand and use this specification well.

Claims

1. A data processing method, characterized in that: include: Collecting hyperspectral region data for a target region, and extracting target region features from the hyperspectral region data; Inputting the target area feature into the object classification model to classify the area object in the target area, and obtaining the object classification information of the area object in the target area, wherein the object classification model optimizes the recognition accuracy of the object classification model for the mixed pixel area through a hierarchical sensitive spectral loss function; Comparing the hyperspectral region data with historical hyperspectral region data of the target region to obtain regional change information of the target region; A visual area report corresponding to the target area is constructed based on the object classification information and the area change information.

2. The data processing method according to claim 1, characterized in that: The step of collecting hyperspectral regional data for the target area includes: The hyperspectral sensor of the UAV is used to collect initial hyperspectral area data for the target area; Preprocessing is performed on the initial hyperspectral region data to obtain hyperspectral region data, wherein the hyperspectral region data has a consistent relationship with the historical hyperspectral region data.

3. The data processing method according to claim 2, characterized in that: The preprocessing of the initial hyperspectral region data to obtain the hyperspectral region data includes: Performing illumination correction processing on the initial hyperspectral region data to obtain intermediate hyperspectral region data; The intermediate hyperspectral region data is subjected to denoising processing by using a Gaussian filter, and the intermediate hyperspectral region data after denoising processing is subjected to viewing angle difference correction processing to obtain target hyperspectral region data; The signal-to-noise ratio of the target hyperspectral region data is calculated, and when it is determined according to the signal-to-noise ratio that the target hyperspectral region data meets a data quality condition, the target hyperspectral region data is used as the hyperspectral region data.

4. The data processing method according to claim 1, characterized in that: Extracting target area features from the hyperspectral area data includes: Performing standardization processing on the feature dimensions of the hyperspectral region data to obtain model output information; The model input information is input into a compression-excitation convolutional neural network for feature extraction to obtain target area features.

5. The data processing method according to claim 1, characterized in that: The step of inputting the target area feature into an object classification model to classify the area object in the target area to obtain the object classification information of the area object in the target area includes: Inputting the target region feature into an object classification model, performing convolution processing on the hyperspectral cube feature corresponding to the region object in the target region feature through a multi-scale convolution kernel in the object classification model, and obtaining a feature map corresponding to the hyperspectral cube feature; Dynamically reweighting the feature map through a compression excitation unit in the object classification model, and determining multi-scale features and multi-level features corresponding to the hyperspectral cube features according to the processing results; The multi-scale features and the multi-level features are spliced ​​into features to be classified, and the features to be classified are processed by a classification unit in the object classification model to obtain object classification information of the regional object in the target area.

6. The data processing method according to any one of claims 1 to 5, characterized in that: The training of the object classification model includes: Obtaining object classification samples, and inputting the object classification samples into an initial object classification model for processing to obtain predicted classification information; Calculating a boundary loss value based on the predicted classification information and the sample classification information corresponding to the object classification sample according to a boundary recognition loss function, and calculating an object loss value based on the predicted classification information and the sample classification information according to an object recognition loss function; Optimizing the initial object classification model based on the boundary loss value and the object loss value until the object classification model that meets the training stop condition is obtained; The boundary recognition loss function and the object recognition loss function constitute the hierarchical sensitive spectrum loss function, and the recognition accuracy of the initial object classification model for object boundaries is optimized by the boundary recognition loss function, and the recognition accuracy of the initial object classification model for regional objects is optimized by the object recognition loss function.

7. The data processing method according to claim 1, characterized in that: Before the step of constructing a visualization area report corresponding to the target area based on the object classification information and the area change information is executed, the step further includes: Determine at least two change regions corresponding to the region change information, and perform spatial aggregation on the at least two change regions to obtain a global change region; Detecting whether the global change region satisfies a region change condition corresponding to the target region; If so, a step of constructing a visualization area report corresponding to the target area based on the object classification information and the area change information is performed.

8. The data processing method according to claim 1, characterized in that: The constructing a visualized area report corresponding to the target area based on the object classification information and the area change information includes: Constructing a classification map corresponding to the target area according to the object classification information; The classification map is updated according to the area change information to obtain an initial change area marking map; The initial change area marking map is smoothed using a preset spatial difference algorithm, and a visualization area report corresponding to the target area is constructed based on the smoothing result.

9. The data processing method according to any one of claims 1 to 5, characterized in that: The step of collecting hyperspectral regional data for the target area includes: Construct a flight strategy for the UAV according to the spectral acquisition range information, spectral resolution information, spatial resolution information, and flight direction attribute information; The UAV is driven based on the flight strategy to collect hyperspectral area data for the target area.

10. A method for classifying urban features, characterized in that: include: Collecting hyperspectral regional data for urban areas, and extracting urban area features from the hyperspectral regional data; Inputting the urban area features into a classification model to classify the ground objects in the urban area, and obtaining ground object classification information of the ground objects in the urban area, wherein the classification model optimizes the recognition accuracy of the classification model for mixed pixel areas through a hierarchical sensitive spectral loss function; Comparing the hyperspectral region data with historical hyperspectral region data of the urban area to obtain regional change information of the urban area; An urban area change report corresponding to the urban area is constructed based on the land feature classification information and the area change information.

11. A data processing device, characterized in that: include: A collection module is configured to collect high-spectral region data for a target region and extract target region features from the high-spectral region data; An input module is configured to input the target area feature into an object classification model to classify the area object in the target area, and obtain object classification information of the area object in the target area, wherein the object classification model optimizes the recognition accuracy of the object classification model for the mixed pixel area through a hierarchical sensitive spectral loss function; A comparison module is configured to compare the hyperspectral region data with historical hyperspectral region data of the target region to obtain region change information of the target region; A construction module is configured to construct a visual area report corresponding to the target area based on the object classification information and the area change information.

12. An urban land feature classification device, characterized in that: include: A data collection module is configured to collect high-spectral area data for urban areas and extract urban area features from the high-spectral area data; An input model module is configured to input the urban area features into a classification model to classify the ground objects in the urban area, and obtain ground object classification information of the ground objects in the urban area, wherein the classification model optimizes the recognition accuracy of the classification model for mixed pixel areas through a hierarchical sensitive spectral loss function; A data comparison module is configured to compare the hyperspectral region data with historical hyperspectral region data of the urban region to obtain regional change information of the urban region; The report building module is configured to build an urban area change report corresponding to the urban area based on the land feature classification information and the area change information.

13. A computing device, characterized in that: include: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the method according to any one of claims 1 to 10 are implemented.

14. A computer-readable storage medium, characterized in that: It stores computer executable instructions, which, when executed by a processor, implement the steps of the method described in any one of claims 1 to 10.

15. A computer program product, characterized in that The method comprises a computer program or an instruction, which, when executed by a processor, implements the steps of the method according to any one of claims 1 to 10.

Images