CN117274584B - Shadow processing method and device for remote sensing image, storage medium and terminal - Google Patents
Shadow processing method and device for remote sensing image, storage medium and terminal Download PDFInfo
- Publication number
- CN117274584B CN117274584B CN202311094600.4A CN202311094600A CN117274584B CN 117274584 B CN117274584 B CN 117274584B CN 202311094600 A CN202311094600 A CN 202311094600A CN 117274584 B CN117274584 B CN 117274584B
- Authority
- CN
- China
- Prior art keywords
- shadow
- model
- feature
- remote sensing
- sensing image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000003672 processing method Methods 0.000 title claims abstract description 16
- 238000013145 classification model Methods 0.000 claims abstract description 127
- 238000012545 processing Methods 0.000 claims abstract description 61
- 238000001514 detection method Methods 0.000 claims abstract description 49
- 238000000034 method Methods 0.000 claims abstract description 31
- 230000004927 fusion Effects 0.000 claims abstract description 28
- 238000001228 spectrum Methods 0.000 claims abstract description 25
- 230000010365 information processing Effects 0.000 claims abstract description 7
- 230000006870 function Effects 0.000 claims description 52
- 238000012549 training Methods 0.000 claims description 38
- 238000002310 reflectometry Methods 0.000 claims description 29
- 238000011156 evaluation Methods 0.000 claims description 25
- 238000009826 distribution Methods 0.000 claims description 24
- 238000004891 communication Methods 0.000 claims description 16
- 238000012937 correction Methods 0.000 claims description 15
- 230000005855 radiation Effects 0.000 claims description 9
- 238000000605 extraction Methods 0.000 description 7
- 238000010276 construction Methods 0.000 description 6
- 238000012216 screening Methods 0.000 description 6
- 230000003595 spectral effect Effects 0.000 description 6
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 5
- 230000008901 benefit Effects 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 239000013598 vector Substances 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000012876 topography Methods 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 238000005286 illumination Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000007499 fusion processing Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000011022 operating instruction Methods 0.000 description 1
- 238000013179 statistical model Methods 0.000 description 1
- 239000000758 substrate Substances 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/273—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion removing elements interfering with the pattern to be recognised
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
- G06V10/765—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects using rules for classification or partitioning the feature space
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
- G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Databases & Information Systems (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a shadow processing method and device of a remote sensing image, a storage medium and a terminal, relates to the technical field of information processing, and mainly aims to solve the problem of low shadow processing accuracy of the remote sensing image. The method mainly comprises the steps of detecting shadows in a target remote sensing image to be processed by using a shadow detection model to obtain a plurality of shadow blocks to be processed, wherein the shadow detection model comprises a spectrum characteristic main model and a multi-characteristic fusion auxiliary model; extracting the characteristics of the shadow block to obtain characteristic variables, and identifying a target classification model matched with the characteristic variables from the trained classification models, wherein the classification model comprises at least one class of characteristic network; and identifying ground object objects in the shadow block by using the target classification model, and reconstructing the shadow block image based on the ground object identification result to obtain a shadow processing result of the target remote sensing image. The method is mainly used for processing shadows in the remote sensing image.
Description
Technical Field
The present invention relates to the field of image processing technologies, and in particular, to a shadow processing method and apparatus for a remote sensing image, a storage medium, and a terminal.
Background
Due to the influence of illumination and topography, shadows usually appear on the high-definition remote sensing image, and the existence of the shadows causes degradation of the visual effect of the remote sensing image, so that the information quantity of the remote sensing image is reduced. Particularly in urban remote sensing images, due to the existence of numerous tall buildings, the ratio of the low-illumination area caused by shadows to the images is high, and the influence caused by shadows is more serious.
The existing shadow processing in the remote sensing image is mainly realized based on shadow detection of spectral characteristics. However, the method has too high prior knowledge requirements on specific image scenes, sunlight conditions and the like, and under the condition of facing more complex scenes, such as urban scenes with more tall buildings, shadows and ground objects have more interweaving and overlapping, so that shadows are difficult to accurately detect and process, and the shadow processing accuracy in a remote sensing image is low.
Disclosure of Invention
In view of the above, the invention provides a shadow processing method and device, a storage medium and a terminal for a remote sensing image, which mainly aim to solve the problem of lower shadow processing accuracy in the existing remote sensing image.
According to one aspect of the present invention, there is provided a shadow processing method for a remote sensing image, including:
Detecting shadows in a target remote sensing image to be processed by using a shadow detection model to obtain a plurality of shadow blocks to be processed, wherein the shadow detection model comprises a spectrum characteristic main model and a multi-characteristic fusion auxiliary model;
Extracting the characteristics of the shadow block to obtain characteristic variables, and identifying a target classification model matched with the characteristic variables from the trained classification models, wherein the classification model comprises at least one class of characteristic network;
And identifying ground object objects in the shadow block by using the target classification model, and reconstructing the shadow block image based on the ground object identification result to obtain a shadow processing result of the target remote sensing image.
Further, before the target classification model matched with the characteristic variable is identified from the classification models which are already trained, the method further comprises:
constructing a plurality of different initial classification models based on feature networks of different feature categories, and acquiring feature variable combination samples matched with the feature categories of the feature networks in each initial classification model;
performing ground object recognition on each initial classification model according to the corresponding characteristic variable sample in the characteristic variable combination sample through a characteristic network in the initial classification model;
Determining a ground object recognition result of the feature variable combination sample based on a single feature recognition result of each feature network and initial weight information of the corresponding feature network;
Constructing a loss function based on the ground object recognition result and the classification precision evaluation result of the ground object recognition result, and training the initial classification model by utilizing the loss function to obtain a trained classification model, wherein the trained classification model comprises a characteristic network corresponding to the weight information of each classification model.
Further, the constructing a loss function based on the feature object recognition result and the classification accuracy evaluation result of the feature object recognition result, and training the initial classification model by using the loss function to obtain a trained classification model, including:
Constructing a first loss function based on the feature object recognition result and the classification label of the feature variable combination sample;
If the result of the first loss function meets a preset first loss function threshold, calculating a classification precision evaluation parameter of the ground object identification result, and constructing a second loss function based on the classification precision evaluation parameter;
And training the initial classification model by using the second loss function to obtain a classification model with the training completed.
Further, the multi-feature fusion auxiliary model is constructed based on a reflectivity sub-model, a shape feature sub-model, a spatial sequence sub-model and a spatial distribution sub-model, and the shadow detection model is used for detecting shadows in a target remote sensing image to be processed to obtain a plurality of shadow blocks to be processed, and the method comprises the following steps:
extracting shadows in the target remote sensing image based on the spectrum characteristic main model to obtain an initial detection result;
Extracting reflectivity, shape characteristics, spatial sequence and spatial distribution of each candidate shadow block in the initial detection result by using the multi-characteristic fusion auxiliary model;
a plurality of shadow blocks to be processed are identified from the candidate shadows based on the reflectivity, the shape features, the spatial sequence, and the spatial distribution.
Further, the reconstructing the shadow block image based on the feature object recognition result to obtain a shadow processing result of the target remote sensing image includes:
Reconstructing an image based on the ground object recognition result to obtain a reconstructed shadow block image;
noise processing and boundary information processing are carried out on the reconstructed shadow block image, and a processed shadow block image is obtained;
and updating a corresponding shadow block in the target remote sensing image based on the processed shadow block image to obtain a shadow processing result of the target remote sensing image.
Further, before detecting the shadow in the target remote sensing image to be processed by using the shadow detection model to obtain a plurality of shadow blocks to be processed, the method further includes:
Acquiring an initial remote sensing image and a reference image;
And performing geometric correction and radiation correction on the initial remote sensing image by using the reference image to obtain a target remote sensing image.
Further, the categories of the feature network include gray features, texture features, and shape features.
According to another aspect of the present invention, there is provided a shadow processing apparatus for remote sensing images, comprising:
The shadow detection module is used for detecting shadows in the target remote sensing image to be processed by using a shadow detection model to obtain a plurality of shadow blocks to be processed, and the shadow detection model comprises a spectrum feature main model and a multi-feature fusion auxiliary model;
The model matching module is used for extracting the characteristics of the shadow block to obtain characteristic variables, and identifying a target classification model matched with the characteristic variables from the trained classification models, wherein the target classification model comprises at least one class of characteristic network;
And the shadow processing module is used for identifying the ground object in the shadow block by utilizing the target classification model, reconstructing the shadow block image based on the ground object identification result, and obtaining a shadow processing result of the target remote sensing image.
Further, the apparatus further comprises:
The construction module is used for constructing a plurality of different initial classification models based on the feature networks of different feature categories, and acquiring feature variable combination samples matched with the feature categories of the feature networks in each initial classification model;
the identification module is used for identifying ground object objects according to the corresponding characteristic variable samples in the characteristic variable combination samples through the characteristic network in the initial classification model aiming at each initial classification model;
The determining module is used for determining the ground object recognition result of the feature variable combination sample based on the single feature recognition result of each feature network and the initial weight information of the corresponding feature network;
The training module is used for constructing a loss function based on the ground object recognition result and the classification precision evaluation result of the ground object recognition result, and training the initial classification model by utilizing the loss function to obtain a classification model which is trained, wherein the classification model which is trained comprises a characteristic network corresponding to the weight information of each classification model.
Further, the training module includes:
The first construction unit is used for constructing a first loss function based on the feature object recognition result and the classification label of the feature variable combination sample;
the second construction unit is used for calculating the classification precision evaluation parameter of the ground object identification result if the result of the first loss function meets a preset first loss function threshold value, and constructing a second loss function based on the classification precision evaluation parameter;
And the training unit is used for training the initial classification model by using the second loss function to obtain a classification model after training.
Further, the shadow detection module includes:
the first extraction unit is used for extracting shadows in the target remote sensing image based on the spectrum characteristic main model to obtain an initial detection result;
The second extraction unit is used for extracting the reflectivity, the shape characteristics, the spatial sequence and the spatial distribution of each candidate shadow block in the initial detection result by utilizing the multi-characteristic fusion auxiliary model;
a plurality of shadow blocks to be processed are identified from the candidate shadows based on the reflectivity, the shape features, the spatial sequence, and the spatial distribution.
Further, the shadow processing module includes:
The reconstruction unit is used for reconstructing an image based on the ground object identification result to obtain a reconstructed shadow block image;
The processing unit is used for carrying out noise processing and boundary information processing on the reconstructed shadow block image to obtain a processed shadow block image;
And the updating unit is used for updating the corresponding shadow block in the target remote sensing image based on the processed shadow block image to obtain a shadow processing result of the target remote sensing image.
Further, the apparatus further comprises:
The acquisition module is used for acquiring an initial remote sensing image and a reference image;
And the correction module is used for carrying out geometric correction and radiation correction on the initial remote sensing image by utilizing the reference image to obtain a target remote sensing image.
Further, in a specific application scenario, the category of the feature network includes gray features, texture features, and shape features.
According to still another aspect of the present invention, there is provided a storage medium having stored therein at least one executable instruction for causing a processor to perform operations corresponding to the shadow processing method of a remote sensing image as described above.
According to still another aspect of the present invention, there is provided a terminal including: the device comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete communication with each other through the communication bus;
The memory is used for storing at least one executable instruction, and the executable instruction enables the processor to execute the operation corresponding to the shadow processing method of the remote sensing image.
By means of the technical scheme, the technical scheme provided by the embodiment of the invention has at least the following advantages:
the embodiment of the invention detects shadows in a target remote sensing image to be processed by utilizing a shadow detection model to obtain a plurality of shadow blocks to be processed, wherein the shadow detection model comprises a spectrum feature main model and a multi-feature fusion auxiliary model; extracting the characteristics of the shadow block to obtain characteristic variables, and identifying a target classification model matched with the characteristic variables from the trained classification models, wherein the classification model comprises at least one class of characteristic network; and identifying the ground object in the shadow block by using the target classification model, reconstructing the shadow block image based on the ground object identification result to obtain a shadow processing result of the target remote sensing image, avoiding the false identification rate of the shadow of the remote sensing image in a complex scene, greatly improving the accuracy of shadow identification in the remote sensing image, and simultaneously training the multi-feature fusion model based on the ground object classification result to improve the accuracy of ground object identification of the shadow block so as to greatly improve the accuracy of shadow processing.
The foregoing description is only an overview of the present invention, and is intended to be implemented in accordance with the teachings of the present invention in order that the same may be more clearly understood and to make the same and other objects, features and advantages of the present invention more readily apparent.
Drawings
Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the invention. Also, like reference numerals are used to designate like parts throughout the figures. In the drawings:
Fig. 1 shows a flowchart of a shadow processing method for a remote sensing image according to an embodiment of the present invention;
FIG. 2 is a flowchart of another method for processing shadows of a remote sensing image according to an embodiment of the present invention;
fig. 3 shows a block diagram of a shadow processing device for remote sensing images according to an embodiment of the present invention;
fig. 4 shows a schematic structural diagram of a terminal according to an embodiment of the present invention.
Detailed Description
Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.
The method is mainly realized based on shadow detection of spectral characteristics aiming at the existing shadow processing in the remote sensing image. However, the method has too high prior knowledge requirements on specific image scenes, sunlight conditions and the like, and under the condition of facing more complex scenes, such as urban scenes with more tall buildings, the shadows and the ground objects have more interweaving and overlapping, so that the shadows are difficult to accurately detect and process, and the problem of lower shadow processing accuracy in the remote sensing image is caused. The embodiment of the invention provides a shadow processing method of a remote sensing image, as shown in fig. 1, comprising the following steps:
101. And detecting shadows in the target remote sensing image to be processed by using the shadow detection model to obtain a plurality of shadow blocks to be processed.
In the embodiment of the invention, the target remote sensing image is a high-definition remote sensing image which needs shadow processing, and can be a remote sensing image of a region where a high-rise building such as an urban remote sensing image, a rural remote sensing image and the like is located. The shadow detection model is used for extracting at least one shadow block in the target remote sensing image and comprises a spectrum characteristic main model and a multi-characteristic fusion auxiliary model. The spectrum characteristic main model is used for extracting spectrum characteristics in the remote sensing image, and primarily judging a region which is likely to be covered by shadow in the image, namely a candidate shadow block, according to the spectrum characteristics of each pixel in the target remote sensing image. The multi-feature fusion auxiliary model acquires other features of the candidate shadow blocks, which are the spectral features of the candidate shadow blocks, and further screens the candidate shadow blocks based on the other features to obtain the shadow blocks to be processed. Wherein the other features may be any two or more of reflectivity, shape features, spatial sequence, spatial distribution. For example, other features include reflectivity and shape features, the multi-feature fusion auxiliary model includes a reflectivity model and a shape feature model, and after a candidate shadow block is obtained, the shape feature is used for extracting shape features and screening the candidate shadow block once to obtain a candidate shadow block after screening once, and then the reflectivity model is used for extracting reflectivity and screening the candidate shadow block after screening once to obtain a shadow block to be processed. The shadow area comprises pixels corresponding to shadow areas and pixels corresponding to shadow surrounding areas.
It should be noted that, detecting shadows according to spectral features is mainly based on the principle that the pixel gray level of a shadow area is darker than that of a non-shadow area, but the situation that the pixel gray level of a water body or other dark substrates is darker is easy to be mistakenly recognized as shadows, further the candidate shadow areas are further recognized by combining the reflectivity, shape features, spatial sequences, spatial distribution and the like of the shadow areas, the reflectivity difference between the shadows and the water body and the shape features and spatial features of shadows caused by a building are fully considered, and more accurate recognition and detection of the shadow areas can be realized, so that the accuracy of subsequent shadow processing is improved.
102. And extracting the characteristics of the shadow block to obtain characteristic variables, and identifying a target classification model matched with the characteristic variables from the trained classification models.
In the embodiment of the invention, the characteristics of the shadow block are extracted. The feature extraction includes both an image basic feature value, such as a gray value, a gray histogram, a mathematical expectation, a frequency, a variance, a covariance, a correlation coefficient, and the like, which are directly extracted from each band image, and a secondary feature, such as a texture analysis result, a ratio processing result, a normalization processing result, a principal component transformation result, and the like, which are obtained by performing an operation processing on the digital image. Because of the different saliency of different features in different images, for example, for highly aggregated building group images, shape features are more pronounced and texture features are more pronounced; aiming at images of parks, lakes and the like in cities with high vegetation coverage and more water bodies, the texture features and the gray features are obvious. Therefore, in order to improve accuracy of feature classification, a classification model matching with the salient features in the shadow block, i.e., a target classification model, is identified from the classification models that have been trained in advance. Wherein the classification model comprises a network of features of at least one category. For example, the classification model may be a model in which a texture feature network and a gray feature network are fused, or may be a model in which a gray feature network and a texture feature network are fused, which is not particularly limited in the embodiment of the present invention.
It should be noted that, the feature network corresponding to different types of features is obtained by pre-training based on different feature samples, and has stronger recognition capability to the corresponding types of features. Meanwhile, the classification model obtained by fusing the corresponding features is selected based on the features in the base shadow block, so that the pertinence of feature recognition can be further improved, and the accuracy of subsequent classification is improved.
103. And identifying ground object objects in the shadow block by using the target classification model, and reconstructing the shadow block image based on the ground object identification result to obtain a shadow processing result of the target remote sensing image.
In the embodiment of the invention, the ground object recognition result is the category of the ground object, such as buildings, trees, water bodies and the like. And reconstructing the shadow block image according to the ground object recognition result and the feature vector so as to eliminate the influence of the shadow on the original features of the ground object and obtain an image which is more attached to the real state. The feature object recognition result in the shadow block is not limited to one, and when a plurality of feature objects exist, image reconstruction is performed on the feature vector and the recognition result of each feature object. The feature vector is a result obtained by further feature space mapping of each feature network to the feature variable. The reconstruction process may implement image reconstruction based on an image generation network, e.g., a semantic-based image generation model.
For further explanation and limitation, in one embodiment of the present invention, before the step of identifying a target classification model matching the feature variable from the classification models that have been trained, as shown in fig. 2, the method further includes:
201. And constructing a plurality of different initial classification models based on the feature networks of different feature categories, and acquiring feature variable combination samples matched with the feature categories of the feature networks in each initial classification model.
202. And aiming at each initial classification model, carrying out ground object recognition according to the corresponding characteristic variable samples in the characteristic variable combination samples through the characteristic network in the initial classification model.
203. And determining the ground object recognition result of the feature variable combination sample based on the single feature recognition result of each feature network and the initial weight information of the corresponding feature network.
204. And constructing a loss function based on the ground object recognition result and the classification accuracy evaluation result of the ground object recognition result, and training the initial classification model by utilizing the loss function to obtain a classification model with completed training.
In an embodiment of the invention, the trained classification model includes a feature network corresponding to the respective weight information. The feature networks of different feature categories are convolutional neural network models which are respectively pre-trained based on corresponding features, feature variables can be further extracted in a feature mode, and classification prediction is carried out on ground features in the image based on feature vectors of single features. And combining the characteristic networks of different characteristic categories to obtain an initial classification model formed by combining the different characteristic networks. And training the initial classification model based on the corresponding feature variable combination samples respectively to determine the weight information of each feature network. For example, training an initial classification model consisting of a shape feature network and a gray feature network by using a feature variable combination sample with obvious shape features and gray features to obtain weight information corresponding to the shape feature network and the gray feature network respectively; and training an initial classification model formed by the shape feature network and the texture feature network by utilizing a feature variable combination sample with obvious shape features and texture features to obtain weight information respectively corresponding to the shape feature network and the texture feature network. The gray feature network is used for processing gray features shown by colors and brightness. And the training process is used for respectively carrying out classification prediction on the corresponding features in the feature variable combination sample based on different feature networks in the initial classification model, calculating to obtain a ground object recognition result according to the obtained sub-classification prediction result and the corresponding initial weight information, and determining a trained loss function by carrying out classification precision evaluation on the ground object recognition result.
In an embodiment of the present invention, for further explanation and limitation, the steps of constructing a loss function based on the feature object recognition result and the classification accuracy evaluation result of the feature object recognition result, and training the initial classification model by using the loss function, to obtain a trained classification model include:
Constructing a first loss function based on the feature object recognition result and the classification label of the feature variable combination sample;
If the result of the first loss function meets a preset first loss function threshold, calculating a classification precision evaluation parameter of the ground object identification result, and constructing a second loss function based on the classification precision evaluation parameter;
And training the initial classification model by using the second loss function to obtain a classification model with the training completed.
In the embodiment of the invention, the initial classification model is trained in two stages, the initial classification model is trained based on the first loss function in the first stage, and when the result of the first loss function is smaller than or equal to the preset first loss function threshold value, the classification precision evaluation is carried out on the classification result trained in the first stage, so that the initial weight information in the initial classification model is further optimized based on the classification precision evaluation parameter, and the training of the model is completed. The first loss function may be a random gradient descent, and the second loss function may be a cross entropy function or other functions, which is not specifically limited in the embodiment of the present invention. The precision evaluation parameter may be an overall accuracy, a user accuracy, a producer accuracy, a Kappa coefficient, etc., and the calculation method may be an error matrix method. Through two-stage training, the model is initially trained based on the first loss function, and tuning is performed based on the classification precision evaluation parameters, so that the accuracy of model training can be improved, and meanwhile, the operation times of the classification precision evaluation parameters are reduced, so that the quality of model training is improved.
In an embodiment of the present invention, for further explanation and limitation, detecting shadows in a target remote sensing image to be processed by using a shadow detection model, obtaining a plurality of shadow blocks to be processed includes:
extracting shadows in the target remote sensing image based on the spectrum characteristic main model to obtain an initial detection result;
Extracting reflectivity, shape characteristics, spatial sequence and spatial distribution of each candidate shadow block in the initial detection result by using the multi-characteristic fusion auxiliary model;
a plurality of shadow blocks to be processed are identified from the candidate shadows based on the reflectivity, the shape features, the spatial sequence, and the spatial distribution.
In the embodiment of the invention, the multi-feature fusion auxiliary model is constructed based on a reflectivity sub-model, a shape feature sub-model, a spatial sequence sub-model and a spatial distribution sub-model. According to the formation mechanism of remote sensing image shadow, the length, direction and shape of the shadow are affected by the light irradiation angle, the light irradiation direction, the topography fluctuation and the like, the shadow of the building often has a regular geometric shape and a certain inclination angle, the shadow also has a certain spatial arrangement relation because of the spatial position relation of the building, and the shadow of the mountain is often consistent with the topography trend. Therefore, after the shadow blocks are initially extracted based on the spectral features to obtain a plurality of candidate shadow blocks, in order to further exclude the erroneously extracted shadow blocks, the reflectivity, the shape features, the spatial sequences and the spatial distributions in each candidate shadow block are respectively extracted based on the reflectivity sub-model, the shape feature sub-model, the spatial sequence sub-model and the spatial distribution sub-model, so that the candidate shadow blocks are further screened based on the reflectivity, the shape features, the spatial sequences and the spatial distributions. Specifically, the screening can be performed based on threshold comparison and feature similarity calculation. For example, filtering out candidate shadow blocks with reflectivity greater than a preset reflectivity threshold as a water body; and filtering out candidate shadow blocks with similarity of at least one of the shape features, the spatial sequences and the spatial distribution with a preset feature sample lower than a similarity threshold value. The false recognition of the shadow can be greatly reduced and the accuracy of the shadow recognition can be improved through the auxiliary screening of the reflectivity, the shape characteristics, the spatial sequence and the spatial distribution multiple characteristics.
In an embodiment of the present invention, for further explanation and limitation, the reconstructing the shadow block image based on the feature object recognition result to obtain a shadow processing result reconstruction of the target remote sensing image, to obtain a shadow processing result of the target remote sensing image, includes:
Reconstructing an image based on the ground object recognition result to obtain a reconstructed shadow block image;
noise processing and boundary information processing are carried out on the reconstructed shadow block image, and a processed shadow block image is obtained;
and updating a corresponding shadow block in the target remote sensing image based on the processed shadow block image to obtain a shadow processing result of the target remote sensing image.
In the embodiment of the invention, after the reconstructed shadow block image is obtained, in order to further improve the processing precision, the shadow block image can be subjected to noise reduction processing and boundary information processing, and the processed image is replaced by the original shadow block image to obtain a shadow processing result. The noise reduction processing may be performed in a time domain or a frequency domain, and since the details of noise and images are distributed in a high-frequency region, the embodiment of the present invention is not particularly limited. Because the signals are mainly distributed in a low-frequency area and the noise is mainly distributed in a high-frequency area, the image noise is reduced, and meanwhile, the image details are required to be kept, so that the purpose of improving the image signal-to-noise ratio is achieved.
In an embodiment of the present invention, for further explanation and limitation, before detecting shadows in the target remote sensing image to be processed by using the shadow detection model to obtain a plurality of shadow blocks to be processed, the method further includes:
Acquiring an initial remote sensing image and a reference image;
And performing geometric correction and radiation correction on the initial remote sensing image by using the reference image to obtain a target remote sensing image.
In the embodiment of the invention, the initial remote sensing image is an image directly acquired by a remote sensing platform. The remote sensing platform can be a satellite remote sensing platform and an unmanned aerial vehicle remote sensing platform, and the embodiment of the invention is not particularly limited. The reference image is the same as the initial remote sensing image shooting object, is corrected and has an accurate reference coordinate system. Since the remote sensing image has geometric distortion, correction of the initial remote sensing image is required. The geometric correction is carried out by adopting a polynomial method, polynomial space transformation and pixel interpolation operation among different images are established through a plurality of control points, registration between the remote sensing image and an actual geographic map piece is realized, and the purposes of correcting geometric changes such as scale, rotation, translation and the like of an initial remote sensing image, reducing and eliminating geometric distortion of the remote sensing image are achieved. The distribution of the repositioned pixels in the original image is uneven, the original image is resampled according to a certain rule according to the position of each pixel on the output image in the input image, interpolation calculation of brightness values is carried out, and a new image matrix is established. The interpolation calculation may be a nearest neighbor interpolation method or a bilinear interpolation method, which is not particularly limited in the embodiment of the present invention. The radiation correction is to correct random radiation distortion or distortion of the system generated by the external factors, the data acquisition and transmission system, and eliminate or correct image distortion caused by radiation errors, and specifically, an image feature model method, a statistical model method or an atmospheric radiation transmission theoretical model method can be used, and the embodiment of the invention is not particularly limited.
In one embodiment of the present invention, for further explanation and limitation, the classes of the feature network include gray features, texture features, shape features.
In the embodiment of the invention, under the condition that the image composition ground feature is single, the image classification effect can be achieved based on single characteristics such as color, shape, texture and the like, but the scenes are only a few of scenes related to the remote sensing image. In reality, the ground feature of a shadow area is often complex, and the image is commonly influenced by various factors. By selecting any two or three of a gray feature network, a texture feature network and a shape feature network to construct a classification model, the feature networks of different feature categories are fused according to different weights, so that multi-feature fusion processing of images is realized, the feature extraction capability of the model on complex scenes and the accuracy of feature recognition are improved, the accuracy of image reconstruction of shadow areas is further improved, and the accuracy of remote sensing image shadow processing is further improved.
The embodiment of the invention provides a shadow processing method of a remote sensing image, which is characterized in that shadows in a target remote sensing image to be processed are detected by utilizing a shadow detection model to obtain a plurality of shadow blocks to be processed, wherein the shadow detection model comprises a spectrum feature main model and a multi-feature fusion auxiliary model; extracting the characteristics of the shadow block to obtain characteristic variables, and identifying a target classification model matched with the characteristic variables from the trained classification models, wherein the classification model comprises at least one class of characteristic network; and identifying the ground object in the shadow block by using the target classification model, reconstructing the shadow block image based on the ground object identification result to obtain a shadow processing result of the target remote sensing image, avoiding the false identification rate of the shadow of the remote sensing image in a complex scene, greatly improving the accuracy of shadow identification in the remote sensing image, and simultaneously training the multi-feature fusion model based on the ground object classification result to improve the accuracy of ground object identification of the shadow block so as to greatly improve the accuracy of shadow processing.
Further, as an implementation of the method shown in fig. 1, an embodiment of the present invention provides a shadow processing apparatus for a remote sensing image, as shown in fig. 3, where the apparatus includes:
The shadow detection module 31 is configured to detect shadows in a target remote sensing image to be processed by using a shadow detection model to obtain a plurality of shadow blocks to be processed, where the shadow detection model includes a spectral feature main model and a multi-feature fusion auxiliary model;
The model matching module 32 is configured to perform feature extraction on the shadow block to obtain a feature variable, and identify a target classification model matched with the feature variable from the classification models that have been trained, where the target classification model includes at least one class of feature network;
and the shadow processing module 33 is configured to identify a ground object in the shadow block by using the target classification model, and reconstruct the shadow block image based on a ground object identification result, so as to obtain a shadow processing result of the target remote sensing image.
Further, the apparatus further comprises:
The construction module is used for constructing a plurality of different initial classification models based on the feature networks of different feature categories, and acquiring feature variable combination samples matched with the feature categories of the feature networks in each initial classification model;
the identification module is used for identifying ground object objects according to the corresponding characteristic variable samples in the characteristic variable combination samples through the characteristic network in the initial classification model aiming at each initial classification model;
The determining module is used for determining the ground object recognition result of the feature variable combination sample based on the single feature recognition result of each feature network and the initial weight information of the corresponding feature network;
The training module is used for constructing a loss function based on the ground object recognition result and the classification precision evaluation result of the ground object recognition result, and training the initial classification model by utilizing the loss function to obtain a classification model which is trained, wherein the classification model which is trained comprises a characteristic network corresponding to the weight information of each classification model.
Further, the training module includes:
The first construction unit is used for constructing a first loss function based on the feature object recognition result and the classification label of the feature variable combination sample;
the second construction unit is used for calculating the classification precision evaluation parameter of the ground object identification result if the result of the first loss function meets a preset first loss function threshold value, and constructing a second loss function based on the classification precision evaluation parameter;
And the training unit is used for training the initial classification model by using the second loss function to obtain a classification model after training.
Further, the shadow detection module 31 includes:
the first extraction unit is used for extracting shadows in the target remote sensing image based on the spectrum characteristic main model to obtain an initial detection result;
The second extraction unit is used for extracting the reflectivity, the shape characteristics, the spatial sequence and the spatial distribution of each candidate shadow block in the initial detection result by utilizing the multi-characteristic fusion auxiliary model;
a plurality of shadow blocks to be processed are identified from the candidate shadows based on the reflectivity, the shape features, the spatial sequence, and the spatial distribution.
Further, the shadow processing module 33 includes:
The reconstruction unit is used for reconstructing an image based on the ground object identification result to obtain a reconstructed shadow block image;
The processing unit is used for carrying out noise processing and boundary information processing on the reconstructed shadow block image to obtain a processed shadow block image;
And the updating unit is used for updating the corresponding shadow block in the target remote sensing image based on the processed shadow block image to obtain a shadow processing result of the target remote sensing image.
Further, the apparatus further comprises:
The acquisition module is used for acquiring an initial remote sensing image and a reference image;
And the correction module is used for carrying out geometric correction and radiation correction on the initial remote sensing image by utilizing the reference image to obtain a target remote sensing image.
Further, in a specific application scenario, the category of the feature network includes gray features, texture features, and shape features.
The embodiment of the invention provides a shadow processing device of a remote sensing image, which is used for detecting shadows in a target remote sensing image to be processed by utilizing a shadow detection model to obtain a plurality of shadow blocks to be processed, wherein the shadow detection model comprises a spectrum characteristic main model and a multi-characteristic fusion auxiliary model; extracting the characteristics of the shadow block to obtain characteristic variables, and identifying a target classification model matched with the characteristic variables from the trained classification models, wherein the classification model comprises at least one class of characteristic network; and identifying the ground object in the shadow block by using the target classification model, reconstructing the shadow block image based on the ground object identification result to obtain a shadow processing result of the target remote sensing image, avoiding the false identification rate of the shadow of the remote sensing image in a complex scene, greatly improving the accuracy of shadow identification in the remote sensing image, and simultaneously training the multi-feature fusion model based on the ground object classification result to improve the accuracy of ground object identification of the shadow block so as to greatly improve the accuracy of shadow processing.
According to one embodiment of the present invention, there is provided a storage medium storing at least one executable instruction for performing the shadow processing method of the remote sensing image in any of the above method embodiments.
Fig. 4 is a schematic structural diagram of a terminal according to an embodiment of the present invention, and the specific embodiment of the present invention is not limited to the specific implementation of the terminal.
As shown in fig. 4, the terminal may include: a processor 402, a communication interface (Communications Interface) 404, a memory 406, and a communication bus 408.
Wherein: processor 402, communication interface 404, and memory 406 communicate with each other via communication bus 408.
A communication interface 404 for communicating with network elements of other devices, such as clients or other servers.
The processor 402 is configured to execute the program 410, and may specifically perform relevant steps in the foregoing embodiment of the shadow processing method for a remote sensing image.
In particular, program 410 may include program code including computer-operating instructions.
The processor 402 may be a central processing unit CPU, or an Application-specific integrated Circuit ASIC (Application SPECIFIC INTEGRATED Circuit), or one or more integrated circuits configured to implement embodiments of the present invention. The one or more processors included in the terminal may be the same type of processor, such as one or more CPUs; but may also be different types of processors such as one or more CPUs and one or more ASICs.
Memory 406 for storing programs 410. Memory 406 may comprise high-speed RAM memory or may also include non-volatile memory (non-volatile memory), such as at least one disk memory.
Program 410 may be specifically operable to cause processor 402 to:
Detecting shadows in a target remote sensing image to be processed by using a shadow detection model to obtain a plurality of shadow blocks to be processed, wherein the shadow detection model comprises a spectrum characteristic main model and a multi-characteristic fusion auxiliary model;
Extracting the characteristics of the shadow block to obtain characteristic variables, and identifying a target classification model matched with the characteristic variables from the trained classification models, wherein the classification model comprises at least one class of characteristic network;
and identifying ground object objects in the shadow block by using the target classification model, and reconstructing the shadow block image based on the ground object identification result to obtain a shadow processing result of the target remote sensing image.
It will be appreciated by those skilled in the art that the modules or steps of the invention described above may be implemented in a general purpose computing device, they may be concentrated on a single computing device, or distributed across a network of computing devices, they may alternatively be implemented in program code executable by computing devices, so that they may be stored in a memory device for execution by computing devices, and in some cases, the steps shown or described may be performed in a different order than that shown or described, or they may be separately fabricated into individual integrated circuit modules, or multiple modules or steps within them may be fabricated into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.
The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (8)
1. A shadow processing method for a remote sensing image, comprising:
Detecting shadows in a target remote sensing image to be processed by using a shadow detection model to obtain a plurality of shadow blocks to be processed, wherein the shadow detection model comprises a spectrum characteristic main model and a multi-characteristic fusion auxiliary model;
Extracting the characteristics of the shadow block to obtain characteristic variables, and identifying a target classification model matched with the characteristic variables from the trained classification models, wherein the classification model comprises at least one class of characteristic network;
Identifying ground object objects in the shadow block by utilizing the target classification model, and reconstructing the shadow block image based on the ground object identification result to obtain a shadow processing result of the target remote sensing image;
The multi-feature fusion auxiliary model is constructed based on a reflectivity sub-model, a shape feature sub-model, a space sequence sub-model and a space distribution sub-model, and the shadow in the target remote sensing image to be processed is detected by using a shadow detection model to obtain a plurality of shadow blocks to be processed, and the multi-feature fusion auxiliary model comprises:
extracting shadows in the target remote sensing image based on the spectrum characteristic main model to obtain an initial detection result;
Extracting reflectivity, shape characteristics, spatial sequence and spatial distribution of each candidate shadow block in the initial detection result by using the multi-characteristic fusion auxiliary model;
Identifying a plurality of shadow blocks to be processed from the candidate shadows based on the reflectivity, the shape features, the spatial sequence, the spatial distribution;
Before the target classification model matched with the characteristic variable is identified from the classification models which are already trained, the method further comprises:
constructing a plurality of different initial classification models based on feature networks of different feature categories, and acquiring feature variable combination samples matched with the feature categories of the feature networks in each initial classification model;
performing ground object recognition on each initial classification model according to the corresponding characteristic variable sample in the characteristic variable combination sample through a characteristic network in the initial classification model;
Determining a ground object recognition result of the feature variable combination sample based on a single feature recognition result of each feature network and initial weight information of the corresponding feature network;
Constructing a loss function based on the ground object recognition result and the classification precision evaluation result of the ground object recognition result, and training the initial classification model by utilizing the loss function to obtain a trained classification model, wherein the trained classification model comprises a characteristic network corresponding to respective weight information;
the spectrum characteristic main model is used for extracting spectrum characteristics in the remote sensing image, and primarily judging a region which is possibly covered by shadow in the image, namely a candidate shadow block, according to the spectrum characteristics of each pixel in the target remote sensing image.
2. The method according to claim 1, wherein the constructing a loss function based on the feature object recognition result and the classification accuracy evaluation result of the feature object recognition result, and training the initial classification model using the loss function, to obtain a trained classification model, comprises:
Constructing a first loss function based on the feature object recognition result and the classification label of the feature variable combination sample;
If the result of the first loss function meets a preset first loss function threshold, calculating a classification precision evaluation parameter of the ground object identification result, and constructing a second loss function based on the classification precision evaluation parameter;
And training the initial classification model by using the second loss function to obtain a classification model with the training completed.
3. The method according to claim 1, wherein reconstructing the shadow block image based on the feature object recognition result to obtain a shadow processing result of the target remote sensing image includes:
Reconstructing an image based on the ground object recognition result to obtain a reconstructed shadow block image;
noise processing and boundary information processing are carried out on the reconstructed shadow block image, and a processed shadow block image is obtained;
and updating a corresponding shadow block in the target remote sensing image based on the processed shadow block image to obtain a shadow processing result of the target remote sensing image.
4. The method of claim 3, wherein before detecting shadows in the target remote sensing image to be processed using the shadow detection model to obtain a plurality of shadow blocks to be processed, the method further comprises:
Acquiring an initial remote sensing image and a reference image;
And performing geometric correction and radiation correction on the initial remote sensing image by using the reference image to obtain a target remote sensing image.
5. The method of any one of claims 1-4, wherein the categories of the feature network include gray features, texture features, shape features.
6. A shadow processing apparatus for a remote sensing image, comprising:
The shadow detection module is used for detecting shadows in the target remote sensing image to be processed by using a shadow detection model to obtain a plurality of shadow blocks to be processed, and the shadow detection model comprises a spectrum feature main model and a multi-feature fusion auxiliary model;
The model matching module is used for extracting the characteristics of the shadow block to obtain characteristic variables, and identifying a target classification model matched with the characteristic variables from the trained classification models, wherein the target classification model comprises at least one class of characteristic network;
the shadow processing module is used for identifying ground object objects in the shadow block by utilizing the target classification model, reconstructing the shadow block image based on the ground object identification result, and obtaining a shadow processing result of the target remote sensing image;
The multi-feature fusion auxiliary model is constructed based on a reflectivity sub-model, a shape feature sub-model, a space sequence sub-model and a space distribution sub-model, and the shadow in the target remote sensing image to be processed is detected by using a shadow detection model to obtain a plurality of shadow blocks to be processed, and the multi-feature fusion auxiliary model comprises:
extracting shadows in the target remote sensing image based on the spectrum characteristic main model to obtain an initial detection result;
Extracting reflectivity, shape characteristics, spatial sequence and spatial distribution of each candidate shadow block in the initial detection result by using the multi-characteristic fusion auxiliary model;
Identifying a plurality of shadow blocks to be processed from the candidate shadows based on the reflectivity, the shape features, the spatial sequence, the spatial distribution;
before the target classification model matched with the characteristic variable is identified from the classification models which are already trained, the method further comprises the following steps:
constructing a plurality of different initial classification models based on feature networks of different feature categories, and acquiring feature variable combination samples matched with the feature categories of the feature networks in each initial classification model;
performing ground object recognition on each initial classification model according to the corresponding characteristic variable sample in the characteristic variable combination sample through a characteristic network in the initial classification model;
Determining a ground object recognition result of the feature variable combination sample based on a single feature recognition result of each feature network and initial weight information of the corresponding feature network;
Constructing a loss function based on the ground object recognition result and the classification precision evaluation result of the ground object recognition result, and training the initial classification model by utilizing the loss function to obtain a trained classification model, wherein the trained classification model comprises a characteristic network corresponding to respective weight information;
the spectrum characteristic main model is used for extracting spectrum characteristics in the remote sensing image, and primarily judging a region which is possibly covered by shadow in the image, namely a candidate shadow block, according to the spectrum characteristics of each pixel in the target remote sensing image.
7. A storage medium having stored therein at least one executable instruction for causing a processor to perform operations corresponding to the shadow processing method of a remote sensing image as defined in any one of claims 1 to 5.
8. A terminal, comprising: the device comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete communication with each other through the communication bus;
The memory is configured to store at least one executable instruction, where the executable instruction causes the processor to perform operations corresponding to the shadow processing method of a remote sensing image according to any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311094600.4A CN117274584B (en) | 2023-08-25 | 2023-08-25 | Shadow processing method and device for remote sensing image, storage medium and terminal |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311094600.4A CN117274584B (en) | 2023-08-25 | 2023-08-25 | Shadow processing method and device for remote sensing image, storage medium and terminal |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117274584A CN117274584A (en) | 2023-12-22 |
| CN117274584B true CN117274584B (en) | 2024-05-03 |
Family
ID=89220477
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311094600.4A Active CN117274584B (en) | 2023-08-25 | 2023-08-25 | Shadow processing method and device for remote sensing image, storage medium and terminal |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117274584B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110186820A (en) * | 2018-12-19 | 2019-08-30 | 河北中科遥感信息技术有限公司 | Multisource data fusion and environomental pollution source and pollutant distribution analysis method |
| CN112580654A (en) * | 2020-12-25 | 2021-03-30 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Semantic segmentation method for ground objects of remote sensing image |
| CN115205618A (en) * | 2022-05-31 | 2022-10-18 | 浙江大华技术股份有限公司 | Earth surface coverage classification model training method, earth surface coverage classification method and device |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110390267B (en) * | 2019-06-25 | 2021-06-01 | 东南大学 | Mountain landscape building extraction method and device based on high-resolution remote sensing image |
-
2023
- 2023-08-25 CN CN202311094600.4A patent/CN117274584B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110186820A (en) * | 2018-12-19 | 2019-08-30 | 河北中科遥感信息技术有限公司 | Multisource data fusion and environomental pollution source and pollutant distribution analysis method |
| CN112580654A (en) * | 2020-12-25 | 2021-03-30 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Semantic segmentation method for ground objects of remote sensing image |
| CN115205618A (en) * | 2022-05-31 | 2022-10-18 | 浙江大华技术股份有限公司 | Earth surface coverage classification model training method, earth surface coverage classification method and device |
Non-Patent Citations (2)
| Title |
|---|
| Fusion of Hyperspectral and LiDAR Data for Classification of Cloud-Shadow Mixed Remote Sensed Scene;Renbo Luo et al.;《IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING》;20170330;第3768-3781页 * |
| 高光谱遥感数据吸收位置特征融合提取技术研究;陶荣华等;《激光杂志》;20131231;第27-28页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117274584A (en) | 2023-12-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110119728B (en) | Remote sensing image cloud detection method based on multi-scale fusion semantic segmentation network | |
| CN111291826B (en) | Pixel-by-pixel classification method of multisource remote sensing image based on correlation fusion network | |
| CN114972191A (en) | Method and device for detecting farmland change | |
| CN110598613B (en) | Expressway agglomerate fog monitoring method | |
| CN113887472B (en) | Remote sensing image cloud detection method based on cascade color and texture feature attention | |
| CN112001403A (en) | Image contour detection method and system | |
| CN112529827A (en) | Training method and device for remote sensing image fusion model | |
| CN113920080A (en) | Power grid fault positioning method based on generation countermeasure network | |
| CN110852207A (en) | Blue roof building extraction method based on object-oriented image classification technology | |
| CN113838064A (en) | Cloud removing method using multi-temporal remote sensing data based on branch GAN | |
| CN116468958A (en) | Communication tower safety detection method and system | |
| CN118429312A (en) | Method, system, equipment and medium for identifying and positioning chip and pin | |
| CN117853942A (en) | Cloud and fog identification method, cloud and fog identification device and cloud and fog identification system | |
| Díaz et al. | Enhanced gap fraction extraction from hemispherical photography | |
| CN117274584B (en) | Shadow processing method and device for remote sensing image, storage medium and terminal | |
| CN113256563A (en) | Method and system for detecting surface defects of fine product tank based on space attention mechanism | |
| CN111179245B (en) | Image quality detection method, device, electronic equipment and storage medium | |
| CN116309213A (en) | High-real-time multi-source image fusion method based on generation countermeasure network | |
| CN115619796A (en) | Method and device for obtaining photovoltaic module template and nonvolatile storage medium | |
| CN115270841A (en) | Bar code detection method and device, storage medium and computer equipment | |
| CN117422654B (en) | Remote sensing image color homogenizing method, device, equipment and storage medium | |
| CN116958833B (en) | GF4 data geometric fine correction method | |
| CN117893413B (en) | Vehicle-mounted terminal man-machine interaction method based on image enhancement | |
| CN117237779B (en) | Image recognition method and system for visible light image and infrared image combined analysis | |
| CN116342417B (en) | Radiation correction method and system for aerial remote sensing image |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |