Mountain shadow terrain classification method and system, electronic equipment and storage medium

Abstract

The invention relates to a mountain shadow terrain classification method and system, electronic equipment and a storage medium, and the classification method comprises the steps: firstly dividing a remote sensing image into a shadow region and a non-shadow region according to the spectral features of different wavebands, and determining the shadow region in the optical remote sensing image; optimizing the shadow region through a terrain correction algorithm and a wavelet transform algorithm; and carrying out ground feature classification identification on the optimized shadow region, and determining a ground feature classification map. The method comprises the following steps: processing a shadow region through a terrain correction algorithm and a wavelet transform algorithm, enhancing the difference of radiance values of different ground features of a remote sensing image of the shadow region, classifying and identifying the ground features in the shadow region, and summarizing to obtain an integral ground feature classification map. According to the scheme, the shadow regions are merged after being classified instead of being integrally classified, so that the ground feature classification accuracy of the whole image is greatly improved.

Description

Technical Field

[0001] The present disclosure belongs to the field of optical remote sensing technology, and in particular relates to a mountain shadow object classification method, system, electronic equipment and storage medium. Background Art

[0002] Land feature classification involves using remote sensing imagery to identify different types of surface cover, such as buildings, water bodies, and vegetation, to better understand and manage the environment. Optical remote sensing imagery, a type of remote sensing technology, acquires surface information through visible and infrared wavelengths. Its rich colors and clear textures make it suitable for land feature identification in a variety of scenarios.

[0003] Shadows in remote sensing imagery are a significant factor affecting the accuracy of object classification. Existing processing methods are all designed to achieve optimal classification results for the entire remote sensing image. However, due to the significant differences in radiation characteristics between shadowed and non-shadowed areas, it is difficult to achieve accurate classification in shadowed areas. Shadowed areas have low reflectivity, which makes their appearance in the image significantly different from that of surrounding non-shadowed areas, resulting in incomplete spectral information. Shadows can lead to misclassification of objects. Because shadows alter the spectral characteristics of objects, spectral-based classification algorithms struggle to accurately distinguish between objects in shadow and different types of objects, which in turn affects the accuracy of classification results. Summary of the Invention

[0004] To address these issues, the present disclosure provides a method, system, electronic device, and storage medium for hillshade object classification. By first dividing remote sensing images into shadow and non-shadow areas, and then specifically optimizing the brightness, contrast, and image detail of shadow areas, the method enhances the differences in radiometric brightness values ​​between different objects in remote sensing images of shadow areas, thereby improving the accuracy of object classification.

[0005] In order to solve the above technical problems, the first aspect of the present invention provides a method for classifying hill shadow objects, the method comprising:

[0006] Acquire an optical remote sensing image within a preset mountain area, and determine shadow areas and non-shadow areas in the optical remote sensing image based on spectral characteristics of different bands;

[0007] Performing ground object classification and identification on the non-shadow area based on a pre-trained supervised classification model to determine a first ground object classification map;

[0008] Optimizing the brightness value of the shadow area by a terrain correction algorithm, and optimizing the contrast and image details of the shadow area by a wavelet transform algorithm;

[0009] Performing ground object classification and identification on the optimized shadow area based on the supervised classification model to determine a second ground object classification map;

[0010] The first land feature classification map and the second land feature classification map are combined to obtain a land feature classification map of the preset mountain area.

[0011] According to a preferred embodiment of the present invention, the optimizing the brightness value of the shadow area by using a terrain correction algorithm includes:

[0012] Obtaining the solar zenith angle and solar incidence angle of the shadow area through the digital elevation model of the preset mountain area and the position of the shadow area;

[0013] Based on the Minnaert terrain correction algorithm, the brightness of each position in the shadow area is optimized by using the Minnaert constant corresponding to the solar zenith angle, the solar incidence angle, and the ground object reflectance characteristics to generate a shadow area with optimized brightness value, contrast, and image details.

[0014] According to a preferred embodiment of the present invention, the optimizing the contrast and image details of the shadow area by using a wavelet transform algorithm includes:

[0015] Processing the shadow area by a wavelet transform algorithm to obtain a low-frequency sub-band and a high-frequency sub-band of the shadow area;

[0016] Adjusting the contrast of the low-frequency sub-band by using a low-frequency sub-band minimum value for maintaining a baseline of image brightness;

[0017] enhancing the image details of the high frequency sub-band by stabilizing the high frequency sub-band average value for detail enhancement;

[0018] Based on the adjusted low-frequency sub-band and the enhanced high-frequency sub-band, a shadow image processed by a wavelet transform algorithm is generated.

[0019] According to a preferred embodiment of the present invention, determining the shadow area and the non-shadow area in the optical remote sensing image according to the spectral characteristics of different bands includes:

[0020] Acquiring spectral features in the optical remote sensing image, wherein the spectral features include: a deep blue band spectrum, a short-wave infrared band spectrum, a near-infrared band spectrum, and a red band spectrum;

[0021] Determine the dark pixel area in the optical remote sensing image according to the deep blue band spectrum and the short infrared band spectrum;

[0022] The shadow area is determined in the dark pixel area according to the near-infrared band spectrum and the red band spectrum through the spectral difference between the water body and the shadow.

[0023] According to a preferred embodiment of the present invention, determining the dark pixel area in the optical remote sensing image based on the deep blue band spectrum and the short infrared band spectrum includes:

[0024] Calculating a first relative deviation between the deep blue band spectrum and the short infrared band spectrum;

[0025] The dark pixel area in the optical remote sensing image is determined according to the magnitude of the first relative deviation and the first preset deviation.

[0026] According to a preferred embodiment of the present invention, determining the shadow area within the dark pixel area by using the spectral difference between the water body and the shadow based on the near-infrared band spectrum and the red band spectrum includes:

[0027] Calculating a second relative deviation between the near-infrared band spectrum and the infrared band spectrum;

[0028] According to the matching result of the second relative deviation value and the second preset deviation interval, the water body and the shadow in the dark pixel area are determined to obtain the shadow area.

[0029] According to a preferred embodiment of the present invention, the classification method further comprises:

[0030] In the shadow area and the non-shadow area, respectively, pixels at the intersection are obtained as mixed pixels;

[0031] The supervised classification model is used to classify the mixed pixels in the shadow area and the non-shadow area respectively, and the type of the ground object of each mixed pixel is determined;

[0032] The ratio of the mixed pixels at the same position in the shadow area and the non-shadow area having the same ground object type is determined. When the ratio is greater than a preset ratio, the ground object classification map meets the requirements.

[0033] In order to solve the above technical problems, the second aspect of the present invention proposes a mountain shadow object classification system based on optical remote sensing images, the classification system comprising:

[0034] A shadow area determination module is used to obtain an optical remote sensing image within a preset mountain area and determine the shadow area and non-shadow area in the optical remote sensing image based on the spectral characteristics of different bands;

[0035] A ground object classification and recognition module is used to classify and recognize ground objects in the non-shadow area based on a pre-trained supervised classification model to determine a first ground object classification map;

[0036] a spectrum optimization module, configured to optimize the brightness value of the shadow area by a terrain correction algorithm, and optimize the contrast and image details of the shadow area by a wavelet transform algorithm;

[0037] The ground object classification and recognition module is further used to perform ground object classification and recognition on the optimized shadow area based on the supervised classification model to determine a second ground object classification map;

[0038] The land feature classification map generating module is used to merge the first land feature classification map and the second land feature classification map to obtain the land feature classification map of the preset mountain area.

[0039] In order to solve the above technical problems, the third aspect of the present invention provides an electronic device, comprising:

[0040] processor; and

[0041] A memory storing computer-executable instructions, wherein when the computer-executable instructions are executed, the processor is caused to perform the method described in any one of the above embodiments.

[0042] In order to solve the above technical problems, the fourth aspect of the present invention proposes a computer storage medium, wherein the computer storage medium stores one or more programs, and when the one or more programs are executed by a processor, the method described in any one of the above embodiments is implemented.

[0043] Compared with the existing technology, the present disclosure has the following advantages: the present disclosure determines the shadow area through the spectral characteristics of different bands in the optical remote sensing image, processes the shadow area through the terrain correction algorithm and the wavelet transform algorithm, optimizes the brightness value, contrast and image details of the shadow area, enhances the difference in the radiation brightness value of different objects in the remote sensing image of the shadow area, and after enhancing the difference between different objects in the shadow area, classifies and identifies the objects in the shadow area through the supervised classification model to generate a ground object classification map of the shadow area, summarizes the ground object classification maps of the optical remote sensing images outside the shadow area to obtain an overall ground object classification map. This solution directly divides the image into shadow area classification and non-shadow area classification, so that the shadow area classification is a new method, and the non-shadow area is directly accurately classified and then spliced, thereby greatly improving the accuracy of the ground object classification of the entire image.

[0044] Other features and advantages of the present disclosure will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present disclosure. The purposes and other advantages of the present disclosure can be realized and obtained by the structures indicated in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present disclosure or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0046] Figure 1 A first flow chart of a hill shadow feature classification method according to an embodiment of the present disclosure is shown;

[0047] Figure 2 A second flowchart of a hill shadow feature classification method according to an embodiment of the present disclosure is shown;

[0048] Figure 3 The third flowchart of a hill shadow feature classification method according to an embodiment of the present disclosure is shown;

[0049] Figure 4 A fourth flowchart of a method for classifying hill shadow objects according to an embodiment of the present disclosure is shown;

[0050] Figure 5 A block diagram of a hill shadow feature classification system according to an embodiment of the present disclosure is shown;

[0051] Figure 6 A schematic structural diagram of an electronic device according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0052] To make the objectives, technical solutions, and advantages of the embodiments of the present disclosure more clear, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present disclosure without making any creative efforts shall fall within the scope of protection of the present disclosure.

[0053] The same reference numerals in the accompanying drawings represent the same or similar elements, components or parts, and thus repeated descriptions of the same or similar elements, components or parts may be omitted below. It should also be understood that although the first, second, third and other numbered adjectives may be used herein to describe various devices, elements, components or parts, these devices, elements, components or parts should not be limited by these adjectives. In other words, these adjectives are only used to distinguish one from another. For example, the first device may also be called the second device, but this does not deviate from the essential technical solution of the present invention. In addition, the terms "and / or" and "and / or" refer to all combinations including any one or more of the listed items.

[0054] See also Figure 1 , Figure 1 This is a flow chart of a hill shadow feature classification method provided by the present invention. Figure 1 As shown, the method includes:

[0055] S11 , obtaining an optical remote sensing image within a preset mountain area, and determining a shadow area and a non-shadow area in the optical remote sensing image according to spectral characteristics of different bands.

[0056] In this embodiment, the optical remote sensing image is an image of the earth's surface captured from space or high altitude by optical sensors. These sensors are usually installed on satellites or aircraft and can receive reflected or radiated electromagnetic waves in the visible light and near-infrared bands from the earth's surface.

[0057] In this embodiment, spectral signatures describe the reflection, absorption, and emission characteristics of ground objects in different wavelength bands. These characteristics are determined by the physical and chemical properties of the ground objects. In remote sensing image analysis, spectral signatures are used to distinguish different ground object types, such as vegetation, water bodies, soil, and buildings.

[0058] In this embodiment, the preset mountain area may be the mountain area on both sides of the river. Due to the terrain of the river and the reflection of the water surface, the features in the shadow area of ​​the mountain area on both sides of the river will be more complex, and it will be more difficult to identify the features in this area.

[0059] In this embodiment, shadows usually appear lower brightness on optical remote sensing images. Therefore, brightness can be used as a feature for shadow detection. By combining the blue and dark appearance of shadows, as well as the reflective properties of vegetation and water bodies, a conversion model can be constructed to suppress features such as vegetation and water bodies while enhancing shadows. Then, the shadow representation is converted into a color space, and candidate shadows are generated using the main color components. The optical remote sensing image can also be converted into an HSV color space. The HSV model is a hexagonal pyramid color space based on hue (H), saturation (S), and value (V). Among them, H represents color information, S represents the purity of the color, and V represents the brightness of the color. In the HSV model, shadow areas usually have a larger hue value H and a lower value V. Therefore, the color RGB image can be converted into an HSV image, and the shadow area can be preliminarily segmented based on these characteristics.

[0060] S12. Classify and identify land features in the non-shadow area based on a pre-trained supervised classification model to determine a first land feature classification map.

[0061] In this embodiment, the image outside the shadow area in the optical remote sensing image is not disturbed by the shadow, and the ground object classification can be performed directly based on the optical remote sensing image.

[0062] In this embodiment, supervised learning algorithms are an important method in machine learning. They use labeled training data to train a model, enabling it to make predictions or decisions about new, unseen data. In a supervised learning algorithm, each training example consists of input variables (usually represented as features) and a desired output variable (also called a label or target). The goal of the algorithm is to learn the mapping between inputs and outputs so that the output of new examples can be accurately predicted. Supervised learning algorithms can be divided into two categories: classification algorithms and regression algorithms.

[0063] In this embodiment, the supervised classification model in this scheme is obtained by training with a supervised learning algorithm. Specifically, multiple groups of historical optical remote sensing images are obtained, and category labels are added to the objects in the historical optical remote sensing images according to the types of objects in the historical optical remote sensing images; based on the category labels of the historical optical remote sensing images and the objects, a supervised classification model is constructed and trained.

[0064] In this embodiment, the images outside the shadow area in the optical remote sensing image are classified and identified by a supervised classification model obtained by pre-training, and the categories of each ground object are determined. Specifically, the images outside the shadow area in the optical remote sensing image are input into the supervised classification algorithm, and the supervised classification algorithm determines the category of each ground object in the image, thereby completing the ground object classification of the optical remote sensing image except the shadow area.

[0065] S13. Optimizing the brightness value of the shadow area by a terrain correction algorithm, and optimizing the contrast and image details of the shadow area by a wavelet transform algorithm.

[0066] In this embodiment, the terrain correction algorithm involves multiple technologies and methods, primarily used in the fields of gravity measurement and remote sensing image processing. Terrain correction refers to eliminating the effects of terrain undulations on observation results through a series of algorithms and techniques, making the observation data more accurate and reliable. Terrain correction in remote sensing images uses various transformations to transform the radiance or reflectance of all pixels to a reference plane (usually a horizontal plane), thereby eliminating image value changes caused by terrain undulations. For example, the band ratio model removes the effect of terrain by simply comparing the radiance or reflectance bands between bands. However, this method fails when the relevant band characteristics of different land objects are similar.

[0067] In this embodiment, the wavelet transform (WT) is a new transform analysis method that inherits and develops the localization concept of the short-time Fourier transform (STFT), while overcoming shortcomings such as the invariance of window size with frequency. It can provide a "time-frequency" window that changes with frequency, making it an ideal tool for signal time-frequency analysis and processing. The basic principle of the wavelet transform is to represent or approximate the original signal using a wavelet function system. The wavelet function system is obtained by translating and scaling the basic wavelet. This function system uses the framework of the function space to obtain the projection of the original signal on this framework, which is the decomposition function of the original signal. Through multi-scale decomposition, the wavelet transform can express signals in the time domain using time scales, thereby achieving excellent processing results. The wavelet transform has a "zoom" property, namely, high frequency resolution and low time resolution (wide analysis window) are used in low-frequency bands, and low frequency resolution and high time resolution (narrow analysis window) are used in high-frequency bands. This property enables the wavelet transform to fully highlight the characteristics of certain aspects of the problem and perform localized analysis of time (space) frequency.

[0068] In this embodiment, the brightness value of the shadow area is adjusted by a terrain correction algorithm. Specifically, for example, Minnaert correction is used to adjust the brightness of a remote sensing image in a mountainous area with uneven illumination according to the nonlinear reflectance characteristics of the ground object.

[0069] In this embodiment, the contrast and image details of the shadow area are adjusted using wavelet transform. Specifically, wavelet transform is widely used in image processing, primarily for image compression and feature extraction. Wavelet multi-resolution decomposition can decompose an image into components of different scales, each with different frequency characteristics. These components of different scales are then adjusted and optimized separately to improve the contrast and image details of the shadow area.

[0070] S14. Classify and identify land features in the optimized shadow area based on the supervised classification model to determine a second land feature classification map; merge the first land feature classification map and the second land feature classification map to obtain a land feature classification map of the preset mountain area.

[0071] In this embodiment, after optimizing the shadow area, the shadow area is classified and identified through a supervised classification model, the categories of each feature in the shadow area are determined to obtain a second feature classification map, and the first feature classification map and the second feature classification map are merged to obtain a feature classification map of the preset mountain area.

[0072] The present invention determines the shadow area through the characteristic features of different bands in the optical remote sensing image, processes the shadow area through a terrain correction algorithm and a wavelet transform algorithm, optimizes the brightness value, contrast and image details of the shadow area, enhances the difference in the radiation brightness value of different objects in the remote sensing image of the shadow area, and after enhancing the difference between different objects in the shadow area, classifies and identifies the objects in the shadow area through a supervised classification model to generate a ground object classification map of the shadow area, summarizes the ground object classification maps of the optical remote sensing images outside the shadow area to obtain an overall ground object classification map, and improves the accuracy of ground object classification.

[0073] In this embodiment, the conformance of the feature classification map can also be determined by the pixel count at the boundary between the optimized shadow and non-shadow areas. Specifically, the classification results are verified by the ratio of overlapping feature types in mixed pixels. Pixels at the boundary between the shadow and non-shadow areas are obtained as mixed pixels. The supervised classification model is then used to classify the mixed pixels in the shadow and non-shadow areas, determining the feature type of each mixed pixel. The ratio of mixed pixels in the same location in the shadow and non-shadow areas that have the same feature type is determined. When this ratio is greater than a preset ratio, the feature classification map meets the requirements. In this solution, "mixed pixels" are obtained that are located on the boundary line between the shadow and non-shadow areas. As an accuracy verification indicator, the ratio of overlapping mixed pixels belonging to the same class after classification in both areas is greater than 98%. The number of overlapping pixels, m, and the number of pixels with consistent classification, n, generally meet the requirements if n / m ≥ 98%.

[0074] See also Figure 2 , Figure 2 This is a second flow chart of a hill shadow feature classification method provided by the present invention. Figure 2 As shown, S13 includes the following steps:

[0075] S21. Obtain the solar zenith angle and solar incidence angle of the shadow area by presetting the digital elevation model of the mountain area and the position of the shadow area.

[0076] In this embodiment, the solar incident angle is the angle between the direct sunlight and the wall normal. The solar zenith angle is the angle between the incident direction of the light and the zenith direction, that is, the angle between the incident light and the ground normal.

[0077] In this embodiment, DEM data is a digital elevation model of the Earth's surface, consisting of point data on the Earth's surface, with corresponding elevation information for each point. This data is usually presented in a raster format, with each grid cell containing an altitude value in meters.

[0078] S22. Based on the Minnaert terrain correction algorithm, the brightness of each position in the shadow area is optimized by using the Minnaert constant corresponding to the solar zenith angle, solar incidence angle, and ground object reflectance characteristics, generating a shadow area with optimized brightness value, contrast, and image details.

[0079] In this embodiment, the Minnaert terrain correction algorithm is based on a non-Lambertian reflectance model. It uses the coefficient k in the Minnaert function to characterize the bidirectional reflectance distribution function (BRDF) of the terrain object, thereby adjusting the intensity of the terrain correction. This algorithm considers the non-Lambertian characteristics of the surface and more closely resembles the actual surface reflectance. Therefore, the correction results are stable and suitable for terrain correction in most mountainous and hilly areas.

[0080] In this embodiment, feature type pre-classification is performed on the image to be corrected, enabling subsequent fitting and solving of the coefficient k for each feature type. Coefficient k fitting and solving is performed on each feature type band by band, obtaining the coefficient k value for each feature type in each band. Terrain correction is performed on the remote sensing image in that range using the coefficient k for each feature type in each band, obtaining the corrected image.

[0081] Specifically, the Minnaert correction is used to adjust the brightness of remote sensing images in mountainous areas with uneven illumination according to the nonlinear reflectance characteristics of the ground objects.

[0082]

[0083] L′(x,y) is the corrected brightness at position (x,y), and L(x,y) is the original image at position (x,y); θ s is the solar zenith angle. i is the angle of incidence, calculated from the DEM data and the observation position. k is the Minnaert constant related to the reflectance characteristics of the ground object.

[0084] See also Figure 3 , Figure 3 The third flow chart of the hill shadow feature classification method provided by the present invention is as follows: Figure 3 As shown, S13 includes the following steps:

[0085] S31. Process the shadow area using a wavelet transform algorithm to obtain a low-frequency sub-band and a high-frequency sub-band of the shadow area.

[0086] In this embodiment, a wavelet transform is performed on the shadow image, typically using a two-dimensional discrete wavelet transform (2D DWT). This process decomposes the image into four sub-bands: a low-frequency sub-band (LL) and three high-frequency sub-bands (LH, HL, HH). Low-frequency sub-band (LL): Contains most of the energy and main structural information of the image, and usually represents the smooth part of the image. High-frequency sub-band: LH (horizontal details): Contains high-frequency information of the image in the horizontal direction, such as edges and textures. HL (vertical details): Contains high-frequency information of the image in the vertical direction. HH (diagonal details): Contains high-frequency information of the image in the diagonal direction.

[0087] S32: Adjust the contrast of the low-frequency sub-band by using the low-frequency sub-band minimum value for maintaining the baseline of the image brightness.

[0088] S33. Enhance the image details of the high-frequency sub-band by using the high-frequency sub-band average value for stabilizing detail enhancement.

[0089] In this embodiment, the overall brightness and contrast of the image are enhanced by adjusting the low-frequency sub-bands obtained by wavelet transform, and the details of the image, especially the edge and texture information, are enhanced by using the high-frequency sub-bands.

[0090] By adjusting the coefficients of the high-frequency subband, image details are enhanced without introducing additional noise.

[0091] I″=α·(I′ LL -min(I′ LL ))+min(I′ LL )+λ·(I′ HF -μ(I′ HF ))

[0092] I″ refers to the image obtained after wavelet transformation enhancement; I′ LL It is the low-frequency subband obtained from wavelet transform, which contains the large-scale structural information of the image. α is the magnification factor for adjusting the contrast, which is adjusted according to the image contrast requirements and is greater than 1. min(I′ LL ) is the minimum value in the low-frequency subband, which is used to maintain the baseline of image brightness. HF It is the high-frequency subband obtained from wavelet transform, which contains the details and edge information of the image. λ is the detail enhancement coefficient, which is between 0 and 1 and is used to control the degree of detail enhancement. HF ) is the average value of the high frequency sub-band, which is used to stabilize the enhancement process and avoid noise caused by over-enhancement.

[0093] S34. Generate a shadow image processed by a wavelet transform algorithm based on the adjusted low-frequency sub-band and the enhanced high-frequency sub-band.

[0094] In this embodiment, after the shadow image is processed by the wavelet transform algorithm, a low-frequency sub-band and a high-frequency sub-band are generated based on the shadow image. Based on the information contained in the low-frequency sub-band and the high-frequency sub-band, corresponding optimization adjustments are performed, and finally a processed shadow image is generated based on the optimized low-frequency sub-band and the high-frequency sub-band.

[0095] See also Figure 4 , Figure 4 This is a flow chart of a hill shadow feature classification method provided by the present invention, as shown in FIG. Figure 4 As shown, S11 includes the following steps:

[0096] S41. Obtain spectral features in optical remote sensing images, including: deep blue band spectrum, shortwave infrared band spectrum, near infrared band spectrum, and red light band spectrum.

[0097] In this embodiment, when direct sunlight cannot reach certain areas on the ground (e.g., mountain shadows), these areas are primarily illuminated by scattered light from the sky. This scattered light is relatively weak, causing these areas to appear darker in the image. By analyzing this scattered light, the location of the shadowed area can be estimated.

[0098] S42. Determine the dark pixel area in the optical remote sensing image based on the deep blue band spectrum and the short infrared band spectrum.

[0099] In this embodiment, the deep blue band (Blue) of the remote sensing image scatters more strongly than the blue light wavelength. Theoretically, the deep blue band (Blue) in the shadow area is higher than the short-wave infrared band (SWIR). Using these two bands for shadow detection makes the features more obvious.

[0100] In this embodiment, dark pixel regions including shadows and water bodies can be determined through spectra of the deep blue band and the short infrared band.

[0101] Specifically, a first relative deviation between the deep blue band spectrum and the short infrared band spectrum is calculated; and the dark pixel area in the optical remote sensing image is determined based on the difference between the first relative deviation and the first preset deviation. The first relative deviation can be a normalized dark pixel index:

[0102]

[0103] NDPI is the normalized dark pixel index, Blue is the dark blue band spectrum, and SWIR is the shortwave infrared band spectrum.

[0104] S43. Determine the shadow area within the dark pixel area based on the spectral difference between the water body and the shadow according to the near-infrared band spectrum and the red light band spectrum.

[0105] Specifically, the second relative deviation between the near-infrared band spectrum and the infrared band spectrum is calculated; based on the matching result of the second relative deviation value and the second preset deviation interval, the water body and shadow in the dark pixel area are determined to obtain the shadow area. The second relative deviation can be the normalized vegetation index. The normalized vegetation index (NDVI) is an index for evaluating vegetation coverage and health status by measuring the difference in reflectance between the near-infrared (NIR) and red (R) bands. The calculation formula is NDVI = (NIR-Red) / (NIR+Red), where NIR and Red represent the reflectance of the near-infrared and red bands, respectively. The value range of NDVI is usually between -1 and 1, where positive values ​​indicate vegetation coverage, negative values ​​indicate water bodies or snow, etc., and values ​​close to 0 indicate rocks or bare soil, etc.

[0106] Furthermore, the above-mentioned normalized dark pixel index and normalized vegetation index can be combined to construct the ratio shadow index (RSI), which can amplify the difference between water bodies and shadows and make it easier to distinguish between shadows and water bodies.

[0107]

[0108] RSI is the ratio shadow index, NDPI is the normalized dark pixel index, and NDVI is the normalized difference vegetation index.

[0109] See also Figure 5 , Figure 5 This is a block diagram of a mountain shadow feature classification system provided by the present invention, such as Figure 5 As shown, the classification system includes: a shadow area determination module 11, a ground object classification recognition module 13, a spectrum optimization module 12 and a ground object classification map generation module 14.

[0110] In this embodiment, the shadow area determination module 11 is used to obtain an optical remote sensing image within a preset mountain area, and determine the shadow area and non-shadow area in the optical remote sensing image according to spectral characteristics of different bands.

[0111] In this embodiment, the ground object classification and recognition module 13 is configured to perform ground object classification and recognition on the non-shadow area based on a pre-trained supervised classification model to determine a first ground object classification map.

[0112] In this embodiment, the spectrum optimization module 12 is configured to optimize the brightness value of the shadow area using a terrain correction algorithm, and to optimize the contrast and image details of the shadow area using a wavelet transform algorithm.

[0113] In this embodiment, the ground object classification and recognition module 13 is further configured to perform ground object classification and recognition on the optimized shadow area based on a supervised classification model to determine a second ground object classification map.

[0114] In this embodiment, the land feature classification map generating module 14 is configured to merge the first land feature classification map and the second land feature classification map to obtain a land feature classification map of a preset mountain area.

[0115] In this embodiment, the spectrum optimization module 12 is specifically used to obtain the solar zenith angle and solar incidence angle of the shadow area by presetting the digital elevation model of the mountain area and the position of the shadow area; based on the Minnaert terrain correction algorithm, the brightness of each position in the shadow area is optimized by using the Minnaert constant corresponding to the solar zenith angle, solar incidence angle and ground reflection characteristics, to generate a shadow area with optimized brightness value, contrast and image details.

[0116] In this embodiment, the spectrum optimization module 12 is specifically configured to process the shadow area using a wavelet transform algorithm to obtain a low-frequency subband and a high-frequency subband of the shadow area; adjust the contrast of the low-frequency subband by using the minimum value of the low-frequency subband for maintaining the baseline of the image brightness; enhance the image details of the high-frequency subband by using the average value of the high-frequency subband for stabilizing detail enhancement; and generate a shadow image processed by the wavelet transform algorithm based on the adjusted low-frequency subband and the enhanced high-frequency subband.

[0117] In this embodiment, the shadow area determination module 11 is specifically used to obtain spectral characteristics in the optical remote sensing image, which spectral characteristics include: a deep blue band spectrum, a short-wave infrared band spectrum, a near-infrared band spectrum, and a red band spectrum; determine the dark pixel area in the optical remote sensing image based on the deep blue band spectrum and the short-wave infrared band spectrum; and determine the shadow area within the dark pixel area based on the spectral difference between the water body and the shadow based on the near-infrared band spectrum and the red band spectrum.

[0118] In this embodiment, the shadow area determination module 11 is specifically used to calculate a first relative deviation between the deep blue band spectrum and the short infrared band spectrum; and determine the dark pixel area in the optical remote sensing image according to the magnitude of the first relative deviation and the first preset deviation.

[0119] In this embodiment, the shadow area determination module 11 is specifically used to calculate the second relative deviation between the near-infrared band spectrum and the infrared band spectrum; based on the matching result of the value of the second relative deviation and the second preset deviation interval, the water body and shadow in the dark pixel area are determined to obtain the shadow area.

[0120] In this embodiment, the classification system also includes: a model training module for obtaining multiple groups of historical optical remote sensing images and adding category labels to the objects in the historical optical remote sensing images; constructing and training a supervised classification model based on the category labels of the historical optical remote sensing images and the objects.

[0121] like Figure 6As shown, an embodiment of the present invention provides an electronic device, including a processor 1110, a communication interface 1120, a memory 1130 and a communication bus 1140, wherein the processor 1110, the communication interface 1120, and the memory 1130 communicate with each other through the communication bus 1140;

[0122] Memory 1130, for storing computer programs;

[0123] The processor 1110 is configured to implement any of the above classification methods when executing the program stored in the memory 1130 .

[0124] In an electronic device provided by an embodiment of the present invention, the processor 1110 obtains an optical remote sensing image within a preset mountain area by executing a program stored on the memory 1130, and determines the shadow area and non-shadow area in the optical remote sensing image according to the spectral characteristics of different bands; classifies and identifies land objects in the non-shadow area based on a pre-trained supervised classification model to determine a first land object classification map; optimizes the brightness value of the shadow area through a terrain correction algorithm, and optimizes the contrast and image details of the shadow area through a wavelet transform algorithm; classifies and identifies land objects in the optimized shadow area based on the supervised classification model to determine a second land object classification map; and merges the first land object classification map and the second land object classification map to obtain a land object classification map of the preset mountain area.

[0125] The communication bus 1140 mentioned in the electronic device can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The communication bus 1140 can be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, the figure shows only one thick line, but this does not mean that there is only one bus or only one type of bus.

[0126] The communication interface 1120 is used for communication between the electronic device and other devices.

[0127] The memory 1130 may include a random access memory 1130 (RAM) or a non-volatile memory 1130, such as at least one disk storage 1130. Alternatively, the memory 1130 may be at least one storage device located away from the processor 1110.

[0128] The above-mentioned processor 1110 can be a general-purpose processor 1110, including a central processing unit 1110 (CPU), a network processor 1110 (NP), etc.; it can also be a digital signal processor 1110 (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0129] An embodiment of the present invention provides a computer-readable storage medium, which stores one or more programs. The one or more programs can be executed by one or more processors 1110 to implement the classification method of any of the above embodiments.

[0130] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware, or any combination thereof. When implemented using software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions according to the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) method. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

[0131] Although the present disclosure has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present disclosure.

Claims

1. A mountain shadow object classification method based on optical remote sensing images, characterized in that: The method comprises: Acquire an optical remote sensing image within a preset mountain area, and determine a shadow area and a non-shadow area in the optical remote sensing image according to spectral characteristics of different bands; Based on the pre-trained supervised classification model, the non-shadow area is classified and identified to determine a first ground object classification map; The brightness value of the shadow area is optimized by a terrain correction algorithm, and the contrast and image details of the shadow area are optimized by a wavelet transform algorithm; Based on the supervised classification model, the optimized shadow area is classified and identified to determine a second land feature classification map; the first land feature classification map and the second land feature classification map are combined to obtain a land feature classification map of the preset mountain area.

2. The classification method according to claim 1, characterized in that: The optimizing the brightness value of the shadow area by using a terrain correction algorithm includes: Obtaining the solar zenith angle and solar incident angle of the shadow area through the digital elevation model of the preset mountain area and the position of the shadow area; Based on the Minnaert terrain correction algorithm, the brightness of each position of the shadow area is optimized by the Minnaert constant corresponding to the solar zenith angle, the solar incidence angle and the reflection characteristics of the ground object, so as to generate a shadow area with optimized brightness value, contrast and image details.

3. The classification method according to claim 2, characterized in that: The optimizing the contrast and image details of the shadow area by using a wavelet transform algorithm comprises: Processing the shadow area by wavelet transform algorithm to obtain a low-frequency sub-band and a high-frequency sub-band of the shadow area; Adjusting the contrast of the low frequency subband by a low frequency subband minimum value for maintaining a baseline of image brightness; The image details of the high frequency sub-band are enhanced by stabilizing the high frequency sub-band average value for detail enhancement; Based on the adjusted low-frequency sub-band and the enhanced high-frequency sub-band, a shadow image processed by a wavelet transform algorithm is generated.

4. The classification method according to claim 1, characterized in that: Determining the shadow area and the non-shadow area in the optical remote sensing image according to the spectral characteristics of different bands includes: Acquire spectral features in the optical remote sensing image, wherein the spectral features include: a deep blue band spectrum, a short-wave infrared band spectrum, a near infrared band spectrum, and a red band spectrum; Determine the dark pixel area in the optical remote sensing image according to the deep blue band spectrum and the short infrared band spectrum; The shadow area is determined in the dark pixel area according to the near infrared band spectrum and the red band spectrum through the spectral difference between the water body and the shadow.

5. The classification method according to claim 4, characterized in that: The step of determining the dark pixel area in the optical remote sensing image according to the deep blue band spectrum and the short infrared band spectrum comprises: Calculating a first relative deviation between the deep blue band spectrum and the short infrared band spectrum; The dark pixel area in the optical remote sensing image is determined according to the magnitude of the first relative deviation and the first preset deviation.

6. The classification method according to claim 4, characterized in that: The determining the shadow area in the dark pixel area by the spectral difference between the water body and the shadow according to the near infrared band spectrum and the red band spectrum includes: Calculating a second relative deviation between the near-infrared band spectrum and the infrared band spectrum; According to the matching result between the value of the second relative deviation and the second preset deviation interval, the water body and the shadow in the dark pixel area are determined to obtain the shadow area.

7. The classification method according to any one of claims 1 to 6, characterized in that: Other classification methods include: In the shadow area and the non-shadow area, respectively, each pixel at the junction is obtained as a mixed pixel; The mixed pixels in the shadow area and the non-shadow area are classified by the supervised classification model to determine the type of the mixed pixels. The ratio of the mixed pixels at the same position in the shadow area and the non-shadow area having the same ground object type is determined. When the ratio is greater than a preset ratio, the ground object classification map meets the requirements.

8. A mountain shadow object classification system based on optical remote sensing images, characterized in that: The classification system includes: A shadow area determination module is used to obtain an optical remote sensing image within a preset mountain area, and determine the shadow area and non-shadow area in the optical remote sensing image according to the spectral characteristics of different bands; A ground object classification and recognition module is used to classify and recognize ground objects in the non-shadow area based on a pre-trained supervised classification model to determine a first ground object classification map; A spectrum optimization module, used to optimize the brightness value of the shadow area by a terrain correction algorithm, and to optimize the contrast and image details of the shadow area by a wavelet transform algorithm; The ground object classification and recognition module is further used to classify and recognize the optimized shadow area based on the supervised classification model to determine a second ground object classification map; The land feature classification map generation module is used to merge the first land feature classification map and the second land feature classification map to obtain the land feature classification map of the preset mountain area.

9. An electronic device, characterized in that: include: processor; as well as A memory storing computer executable instructions which, when executed, cause the processor to perform the method according to any one of claims 1-7.

10. A computer storage medium, characterized in that: in, The computer storage medium stores one or more programs, and when the one or more programs are executed by a processor, the method of any one of claims 1 to 7 is implemented.

Images