Ice and snow melting infiltration detection method and device, medium and equipment

Abstract

The invention discloses an ice and snow ablation infiltration detection method and device, a medium and equipment, and the method comprises the steps: obtaining a remote sensing image of a target region, inputting the remote sensing image into a plurality of trained different accumulated snow classification models, and obtaining a classification result of each accumulated snow classification model; performing fusion processing on the classification result of each accumulated snow classification model, and calculating the accumulated snow coverage area of the target area according to the fusion result; and constructing a relation model of the melt infiltration amount and the ice and snow ablation amount, and calculating the melt infiltration amount of the target area based on the accumulated snow coverage area of the target area and the relation model of the melt infiltration amount and the ice and snow ablation amount. As the classification results of the plurality of accumulated snow classification models are fused, the accuracy of the fused classification result is relatively high, the obtained accumulated snow coverage area is relatively accurate, and meanwhile, the relation model between ice and snow ablation and the infiltration water amount considers the complex condition of ice and snow ablation of the strip mine in the cold region; therefore, the accuracy of the calculated melt water infiltration amount is high.

Description

Technical Field

[0001] The present invention relates to the technical field of ice and snow melting, and in particular to a method, device, medium and equipment for detecting ice and snow melting and infiltration. Background Art

[0002] In high-altitude and cold regions, mining faces extremely severe challenges from natural conditions, especially during the snow accumulation in winter and the melting of ice and snow in spring, when the seepage field of the mine will change significantly. With the increasing depth of mining and the change of seasons, coupled with the impact of mining disturbance, the original groundwater seepage field may be redistributed, greatly increasing the risk of slope seepage and water gushing. In addition, due to the complexity of the properties of the slope rock and soil and the diversity of influencing factors, accompanied by dangerous factors such as snow seepage, mining activities in high-altitude and cold open-pit mining projects may cause slope landslides, especially as the mining depth of the mine increases, the safety problem of high and steep slopes becomes more prominent. Therefore, it is extremely important to conduct infiltration detection of ice and snow meltwater in the mining area.

[0003] At present, the detection of ice and snow meltwater infiltration in high-altitude open-pit mines mainly relies on satellite remote sensing technology. Satellite images are obtained to obtain the reflectivity data of ice and snow in the visible light band. As the ice and snow melt, their surface becomes rough, impurities increase, and the reflectivity changes. By comparing the reflectivity data at different times, the process of ice and snow melting and infiltration can be monitored.

[0004] However, there are some interference factors on the surface of the mining area, and the reflectivity obtained cannot accurately reflect the snow accumulation status. Therefore, the existing mining area ice and snow melting detection methods are difficult to accurately detect the ice and snow melting infiltration situation. Summary of the invention

[0005] In view of this, the present invention provides a method, device, medium and equipment for detecting ice and snow melting and infiltration, the main purpose of which is to solve the problem that the current method for detecting ice and snow melting in mining areas is difficult to accurately detect ice and snow melting and infiltration.

[0006] According to one aspect of the present application, a method for detecting ice and snow melting and infiltration is provided, the method comprising: Acquire a remote sensing image of the target area, and input the remote sensing image into a plurality of different snow classification models that have been trained to obtain a classification result of each snow classification model, wherein the classification result includes the confidence of a plurality of category labels corresponding to each pixel in the remote sensing image; The classification results of each snow classification model are fused, and the snow cover area of ​​the target area is calculated according to the fusion results; Constructing a relationship model between meltwater infiltration and ice and snow melting, and calculating the meltwater infiltration in the target area based on the snow cover area of ​​the target area and the relationship model between meltwater infiltration and ice and snow melting; The step of fusing the classification results of each snow classification model and calculating the snow cover area of ​​the target area according to the fusion results includes: For each of the snow classification models, the confidence corresponding to each category is obtained, and the confidence of the category label corresponding to each pixel is multiplied by the confidence corresponding to the category label to obtain the confidence distribution data corresponding to each pixel; Calculate the conflict factor based on the trust distribution data corresponding to each pixel point of each snow classification model and a preset conflict factor calculation formula; Based on the conflict factor, the trust distribution data corresponding to each pixel of each snow classification model and a preset trust fusion calculation formula, the fused trust distribution data of each pixel is calculated; Based on the fused trust distribution data of each pixel, the comprehensive category of the pixel is determined, and the product of the number of pixels with the comprehensive category of snow and the preset image spatial resolution is used as the snow coverage area of ​​the target area.

[0007] Optionally, before obtaining the confidence corresponding to each category for each snow classification model, the method for detecting ice and snow melting and infiltration further includes: Obtain a validation data set corresponding to each snow classification model, and input the validation data set into each trained snow classification model to obtain a model classification result corresponding to each snow classification model; Perform positive and negative statistics of the classification results based on the actual classification results of the validation data set and the model classification results, and generate a confusion matrix corresponding to each of the snow classification models according to the statistical results; The confidence corresponding to each category of each snow classification model is obtained by performing calculation based on the confusion matrix corresponding to each snow classification model and a preset confidence calculation formula.

[0008] Optionally, the step of constructing a relationship model between meltwater infiltration and ice and snow melting includes: The snow sublimation model was constructed using the surface temperature method; Based on the balance relationship between the total amount of ice and snow, the amount of ice and snow melt, and the surface runoff and the snow sublimation model, a relationship model between meltwater infiltration and ice and snow melt is constructed.

[0009] Optionally, the relationship model between the meltwater infiltration and the ice and snow melting is: , in, R is the meltwater infiltration; is the snow density; S n is the snow area; h is the average thickness of snow; I is the surface runoff, For evaporation, R n is the net radiation, S Energy stored in snow and soil, is the air density, C p is the specific heat of air at constant pressure, f(u x ) is the ground energy transfer equation, ( T s -T a ) is the temperature difference between the snow surface and the air.

[0010] Optionally, before inputting the remote sensing image into a plurality of trained snow classification models to obtain the classification results of each snow classification model, the snow melting and infiltration detection method further includes: Acquire multiple original remote sensing images of the training area, and determine the snow-covered area in each original remote sensing image; Intercepting each of the original remote sensing images according to a preset image size to obtain a plurality of training images, and determining the snow classification category and snow coverage area of ​​each training image; Based on the multiple training images and their corresponding snow classification categories and snow-covered areas, multiple different initial snow classification models are trained respectively to obtain multiple trained snow classification models.

[0011] Optionally, the multiple snow classification models include snow classification machine learning models constructed using K-nearest neighbor algorithm, random forest algorithm, support vector machine algorithm and adaptive boosting algorithm respectively.

[0012] According to another aspect of the present application, a device for detecting ice and snow melting and infiltration is provided, comprising: A classification module, used to obtain a remote sensing image of a target area, and input the remote sensing image into a plurality of trained snow classification models to obtain a classification result of each snow classification model, wherein the classification result includes the confidence of a plurality of category labels corresponding to each pixel in the remote sensing image; A fusion module is used to fuse the classification results of each snow classification model and calculate the snow cover area of ​​the target area according to the fusion results; A calculation module is used to construct a relationship model between meltwater infiltration and ice and snow melting, and calculate the meltwater infiltration of the target area based on the snow cover area of ​​the target area and the relationship model between meltwater infiltration and ice and snow melting; Wherein, the fusion module is also used for: For each of the snow classification models, the confidence corresponding to each category is obtained, and the confidence of the category label corresponding to each pixel is multiplied by the confidence corresponding to the category label to obtain the confidence distribution data corresponding to each pixel; Calculate the conflict factor based on the trust distribution data corresponding to each pixel point of each snow classification model and a preset conflict factor calculation formula; Based on the conflict factor, the trust distribution data corresponding to each pixel of each snow classification model and a preset trust fusion calculation formula, the fused trust distribution data of each pixel is calculated; Based on the fused trust distribution data of each pixel, the comprehensive category of the pixel is determined, and the product of the number of pixels with the comprehensive category of snow and the preset image spatial resolution is used as the snow coverage area of ​​the target area.

[0013] Optionally, the ice and snow melting and infiltration detection device further comprises: A category trust acquisition module is used to obtain a verification data set corresponding to each snow classification model, and input the verification data set into each trained snow classification model to obtain a model classification result corresponding to each snow classification model; perform positive and negative statistics on the classification results based on the actual classification results of the verification data set and the model classification results, and generate a confusion matrix corresponding to each snow classification model according to the statistical results; perform calculations based on the confusion matrix corresponding to each snow classification model and a preset trust calculation formula to obtain the trust corresponding to each category of each snow classification model.

[0014] Optionally, the calculation module is further used for: The snow sublimation model was constructed using the surface temperature method; Based on the balance relationship between the total amount of ice and snow, the amount of ice and snow melt, and the surface runoff and the snow sublimation model, a relationship model between meltwater infiltration and ice and snow melt is constructed.

[0015] Optionally, the relationship model between the meltwater infiltration and the ice and snow melting is: , in, R is the meltwater infiltration; is the snow density; S n is the snow area; his the average thickness of snow; I is the surface runoff, For evaporation, R n is the net radiation, S Energy stored in snow and soil, is the air density, C p is the specific heat of air at constant pressure, f(u x ) is the ground energy transfer equation, ( T s -T a ) is the temperature difference between the snow surface and the air.

[0016] Optionally, the ice and snow melting and infiltration detection device further comprises: A training module is used to obtain multiple original remote sensing images of a training area and determine the snow-covered area in each original remote sensing image; to intercept each of the original remote sensing images according to a preset image size to obtain multiple training images, and to determine the snow classification category and snow-covered area of ​​each training image; and to train multiple different initial snow classification models based on the multiple training images and their corresponding snow classification categories and snow-covered areas to obtain multiple trained snow classification models.

[0017] Optionally, the multiple snow classification models include snow classification machine learning models constructed using K-nearest neighbor algorithm, random forest algorithm, support vector machine algorithm and adaptive boosting algorithm respectively.

[0018] According to another aspect of the present application, a storage medium is provided, in which at least one executable instruction is stored, and the executable instruction enables a processor to execute operations corresponding to the above-mentioned method for detecting ice and snow melting and infiltration.

[0019] According to another aspect of the present application, a computer device is provided, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute operations corresponding to the above-mentioned ice and snow melting and infiltration detection method.

[0020] By means of the above technical solution, the technical solution provided by the embodiment of the present invention has at least the following advantages: The present application provides a method, device, equipment and medium for detecting snow melting and infiltration. Snow classification detection is performed on remote sensing images of a target area through multiple different snow classification models to obtain multiple classification results. The multiple classification results are fused, and the snow cover area is calculated according to the fused classification results. The calculated snow cover area is input into a relationship model between meltwater infiltration and snow melting to calculate the meltwater infiltration of the target area. Since the results of multiple snow classification models are fused, the classification results after fusion are more accurate, the obtained snow cover area is more accurate, and the accuracy of meltwater infiltration is improved. At the same time, the relationship model between snow melting and infiltration water takes into account the complex conditions of snow melting in open-pit mines in cold regions, and also improves the accuracy of meltwater infiltration.

[0021] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings: Figure 1 A flow chart of a method for detecting ice and snow melting and infiltration provided by an embodiment of the present application is shown; Figure 2 A flow chart showing another method for detecting ice and snow melting and infiltration provided in an embodiment of the present application is shown; Figure 3 The image shows the snow classification result output by multiple different snow classification models of a method for detecting snow melting and infiltration provided by an embodiment of the present application; Figure 4 A schematic diagram showing a method for detecting ice and snow melting and infiltration provided by an embodiment of the present application, which integrates snow classification results output by multiple different snow classification models; Figure 5 An image showing a method for detecting ice and snow melting and infiltration provided by an embodiment of the present application after fusing snow classification results output by multiple different snow classification models; Figure 6 A comparison diagram showing a snow cover classification result obtained by photographing a real mining area with a drone in a method for detecting ice and snow melting and infiltration provided by an embodiment of the present application and a snow classification result after fusion; Figure 7 A block diagram showing a component of a device for detecting ice and snow melting and infiltration provided by an embodiment of the present application is shown; Figure 8 A schematic structural diagram of a computer device provided by an embodiment of the present invention is shown.

[0023] in, Figure 7 In: 702-classification module; 704-fusion module; 706-computation module; Figure 8 In it: 802 - processor; 804 - communication interface; 806 - memory; 808 - communication bus; 810 - program. DETAILED DESCRIPTION

[0024] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present invention can be combined with each other without conflict.

[0025] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects of the present invention application are described in detail below in conjunction with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures, or characteristics in one or more embodiments may be combined in any suitable form.

[0026] In view of the problem that the current snow and ice melting detection method in mining areas is difficult to accurately detect the infiltration of snow and ice melting, the embodiment of the present application provides a method for detecting the infiltration of snow and ice melting, such as Figure 1 As shown, the method includes: 102: Obtain a remote sensing image of the target area, and input the remote sensing image into a plurality of trained different snow classification models to obtain a classification result of each snow classification model; 104: Fusing the classification results of each snow classification model, and calculating the snow cover area of ​​the target area according to the fusion results; 106: Construct a relationship model between meltwater infiltration and snow and ice melting. Based on the snow cover area of ​​the target area and the relationship model between meltwater infiltration and snow and ice melting, calculate the meltwater infiltration in the target area.

[0027] Specifically, the target area is the open-pit mine slope. High-resolution optical remote sensing images of the snow-covered area of ​​the open-pit mine slope are obtained, and the remote sensing images are preprocessed, resampled, and panchromatic sharpening are performed on the remote sensing images. The preprocessed remote sensing images are then input into multiple different trained snow classification models, such as the snow classification models established using the K nearest neighbor algorithm, random forest algorithm, support vector machine algorithm, and adaptive boosting algorithm. These different snow classification models are used to classify the preprocessed remote sensing images into snow and non-snow areas.

[0028] The Dempster-Shafer framework is used to fuse the classification results output by multiple different snow classification models to obtain the fused classification results, and the snow cover area is calculated based on the fused classification results. Based on environmental conditions, topography and other parameters, a relationship model between meltwater infiltration and snow and ice melting is constructed on the basis of the heat conduction equation. The calculated snow cover area is substituted into the relationship model between meltwater infiltration and snow and ice melting to calculate the meltwater infiltration in the target area.

[0029] The present application provides a method for detecting snow and ice melting infiltration. Compared with the prior art, a plurality of different snow classification models are used to perform snow classification detection on remote sensing images of a target area to obtain a plurality of classification results, which are fused and processed, and the snow cover area is calculated according to the fused classification results. The calculated snow cover area is input into a relationship model between meltwater infiltration and snow and ice melting to calculate the meltwater infiltration of the target area. Since the results of a plurality of snow classification models are fused, the classification results after fusion are more accurate, the obtained snow cover area is more accurate, and the accuracy of meltwater infiltration is improved. At the same time, the relationship model between snow and ice melting and infiltration water volume takes into account the complex conditions of snow and ice melting in open-pit mines in cold regions, and also improves the accuracy of meltwater infiltration.

[0030] In one embodiment of the present invention, Figure 2 As shown in the figure, the classification results include the confidence of multiple category labels corresponding to each pixel in the remote sensing image; the classification results of each snow classification model are fused, and the snow cover area of ​​the target area is calculated based on the fusion results, including: 202: for each snow classification model, obtain the confidence corresponding to each category, multiply the confidence of the category label corresponding to each pixel by the confidence corresponding to the category label, and obtain the confidence distribution data corresponding to each pixel; 204: Calculate a conflict factor based on the trust distribution data corresponding to each pixel point of each snow classification model and a preset conflict factor calculation formula; 206: Calculate fusion trust distribution data for each pixel based on the conflict factor, the trust distribution data corresponding to each pixel of each snow classification model, and a preset trust fusion calculation formula; 208: Based on the fused trust distribution data of each pixel, determine the comprehensive category of the pixel, and use the product of the number of pixels with the comprehensive category of snow and the preset image spatial resolution as the snow coverage area of ​​the target area.

[0031] In this embodiment, the combination rules of the DS (Dempster-Shafer) framework are used to fuse these different classification results, so as to obtain a more reliable classification result. The classification results of different snow classification models for the same remote sensing image are obtained, and the trust distribution data corresponding to each pixel in the classification result of each snow classification model is calculated. Taking a pixel in the classification result of a snow classification model as an example, the trust of the snow classification model is obtained based on a preset verification data set, and the confidence of the category label corresponding to the pixel in the classification result is multiplied by the trust of the category corresponding to the category label, and the product is used as the trust distribution data corresponding to the pixel. The same method is used to calculate the trust distribution data corresponding to each pixel in the classification result of each snow classification model.

[0032] Then, the DS fusion rule is used to combine the trust distribution data corresponding to each pixel in the classification results of different snow classification models to obtain comprehensive fused trust distribution data. Specifically, the conflict factor corresponding to each pixel is first calculated. After obtaining the conflict factor, the conflict factor corresponding to each pixel and the trust distribution data corresponding to each pixel of each snow classification model are substituted into the preset trust fusion calculation formula to calculate the fused trust distribution data of each pixel.

[0033] Taking two different snow classification models as an example, the BPA (trust distribution data) of each snow classification model for pixel P is: , in, and are the confidence of snow classification model i for “snow” and “non-snow” respectively, is the confidence of snow classification model i in the uncertainty of classification results.

[0034] Based on the trust distribution data of each pixel in the classification results of each snow classification model, the conflict factor corresponding to each pixel is calculated. The conflict factor K represents the inconsistency between classifiers, and the calculation formula is: , in, is the confidence of the first snow classification model on “snow”, is the trust of the second snow classification model on “non-snow”. For the two snow classification models M1 and M2, the conflict factor is: , in, is the confidence that the pixel is snow in the first snow classification model, is the confidence that the pixel is not snowy in the first snow classification model, is the confidence that the pixel is snow in the second snow classification model, is the confidence that the pixel is not snowy in the second snow classification model.

[0035] Based on the Dempster-Shafer framework, the classification results of different classification models are integrated, and the Dempster combination rule is used to calculate the fused BPA: , For two snow classification models M1 and M2, the fused BPA is: , in, is the confidence of the first snow classification model on the uncertainty of the classification result, is the confidence of the second snow classification model on the uncertainty of the classification result.

[0036] If there are multiple classification models, use Dempster combination rules to merge them one by one. For example: First merge M1 and M2 to get M 12 , and then merge M 12 and M3, we get M 123 , and so on, until the BPA fusion of all classification models is completed, and the fusion trust distribution data of the pixel is obtained. According to the fusion trust distribution data of the pixel, the comprehensive category of the pixel is determined, and the category with the highest trust is usually selected as the final comprehensive category. The above method is used to determine the comprehensive category of each pixel.

[0037] In one embodiment of the present invention, for each snow classification model, before obtaining the confidence corresponding to each category, the method for detecting snow and ice melting infiltration further includes: Obtain the validation data set corresponding to each snow classification model, input the validation data set into each trained snow classification model, and obtain the model classification result corresponding to each snow classification model; Based on the actual classification results of the validation data set and the model classification results, the positive and negative statistics of the classification results are performed, and the confusion matrix corresponding to each snow classification model is generated according to the statistical results; The confidence corresponding to each category of each snow classification model is obtained based on the confusion matrix corresponding to each snow classification model and the preset confidence calculation formula.

[0038] Specifically, a validation data set is obtained while obtaining a training data set. Taking a snow classification model as an example, after the snow classification model is trained, the validation data set is input into the trained snow classification model to obtain the model classification result corresponding to the validation data set. The model classification result corresponding to the validation data set is compared with the actual classification result to determine the true or false of each validation sample. For example, (TP) = true positive example, a sample correctly predicted to be a positive class, (FN) = false negative example, a sample that is actually a positive class but predicted to be a negative class, (FP) = false positive example, a sample that is actually a negative class but predicted to be a positive class, (TN) = true negative example, a sample correctly predicted to be a negative class, determine the number of true and false samples, and generate a confusion matrix corresponding to the snow classification model.

[0039] For each snow classification model, the confidence level of each snow classification model in different categories is calculated according to the confusion matrix corresponding to the snow classification model. C i , the confidence of the classification model can be calculated by the following formula: , in, TP i and TN i The categories are C i The number of true positive examples and the number of true negative examples, FP i and FN i The categories are C i The number of false positives and false negatives.

[0040] The uncertainty of the classification model can be estimated by the error rate: , The above method is used to calculate the trust corresponding to each category of each snow classification model.

[0041] In one embodiment of the present invention, a relationship model between meltwater infiltration and ice and snow melting is constructed, including: The snow sublimation model was constructed using the surface temperature method; Based on the balance relationship between the total amount of ice and snow, the amount of ice and snow melt, and the surface runoff and the snow sublimation model, a relationship model between meltwater infiltration and ice and snow melt is constructed.

[0042] Specifically, the amount of meltwater infiltration is expressed as the product of snow density, snow area and average snow thickness, and is associated with snow sublimation and surface runoff. Combining environmental parameters such as snow density, snow area, average snow thickness, surface runoff, net radiation, energy storage, air density, latent heat of sublimation, air specific heat, surface wind speed and temperature difference between snow surface and air, and taking into full consideration the complex conditions of snow and ice melting in open-pit mines in cold regions, the energy balance model is used to quantitatively model the relationship between snow and ice melting and infiltration.

[0043] By collecting information on the climate of the study area, the surface temperature method is used to construct an ice and snow evaporation model. The surface temperature method is based on the temperature difference between the air and the snow surface, combined with the ground energy transfer equation, to calculate the sensible heat flux of the snow surface. By bringing the sensible heat flux into the surface energy balance equation, the latent heat of snow evaporation can be indirectly estimated, thereby calculating the amount of ice and snow sublimation. The snow sublimation model describes the process of snow turning into water vapor, in which the amount of snow sublimation is related to net radiation, energy storage, air density, latent heat of sublimation, specific heat of air, surface wind speed, and the temperature difference between the snow surface and the air.

[0044] The snow sublimation model was initially constructed using the surface temperature method: , In the formula, is the latent heat of evaporation (W.kg -1 ); E is the amount of snow sublimation (kg.m -2 ); R n is the net radiation (Wm -2 ); S is the energy stored in snow and soil (Wm -2 ); is the air density (kg.m -3 ); C p is the specific heat of air at constant pressure (J / kg.℃); f(u x ) is the ground energy transfer equation; ( T s -T a ) is the temperature difference between the snow surface and the air (℃).

[0045] The relationship model between meltwater infiltration and ice and snow melting is: , in, R is the meltwater infiltration; is the snow density; S n is the snow area; h is the average thickness of snow; I is the surface runoff, For evaporation, R n is the net radiation, S Energy stored in snow and soil, is the air density, C p is the specific heat of air at constant pressure, f(u x ) is the ground energy transfer equation, ( T s -T a ) is the temperature difference between the snow surface and the air.

[0046] In one embodiment of the present invention, before inputting the remote sensing images into a plurality of trained snow classification models and obtaining the classification results of each snow classification model, the method for detecting snow melting and infiltration further includes: Acquire multiple original remote sensing images of the training area and determine the snow covered area in each original remote sensing image; Each original remote sensing image is intercepted according to a preset image size to obtain a plurality of training images, and the snow classification category and snow coverage area of ​​each training image are determined; Based on multiple training images and their corresponding snow classification categories and snow-covered areas, multiple different initial snow classification models are trained respectively to obtain multiple trained snow classification models.

[0047] Specifically, high-resolution optical remote sensing images of snow-covered open-pit mine slopes are obtained as original remote sensing images, and preprocessing, resampling, and panchromatic sharpening are performed on the original remote sensing images. Then, the normalized snow index method is combined with the manual confirmation method to determine the snow coverage area in the original remote sensing image, which serves as the true value of the training set for subsequent model training.

[0048] The preprocessing operations of remote sensing spectral images include atmospheric correction to eliminate the influence of the atmosphere on the light signal received by the sensor, geometric correction to eliminate the image of the terrain, and radiation correction to convert the numerical values ​​received by the sensor into radiance values. In addition, in multispectral images, the spectral resolution of the spectral band is higher than that of the spatial band, while the spatial resolution of the panchromatic band is higher than that of the spectral band. The panchromatic sharpening method is used, and the multispectral band and the spatial band are merged on the basis of the panchromatic band. The image processed by panchromatic sharpening can identify land features with high accuracy. In the panchromatic sharpening method, in order to maintain the quality of spatial information and spectral information, the local mean and variance matching (LMVM) algorithm is used. The local mean and variance matching algorithm is a technique used in image processing that can adjust the local contrast and brightness distribution of the image. This method measures the local mean and variance of the image and then adjusts the image according to the desired style or characteristics. Through local mean and variance matching, the visual quality of the image can be improved, the details of the image can be enhanced, and the image can be made clearer and easier to observe. Because the research area is a mining area with complex terrain, the illumination difference between the shady slope and the sunny slope of the remote sensing image is large. Terrain correction refers to the use of different conversion methods to map the radiation brightness of all pixels to a standard plane to reduce the impact of terrain undulations on the image radiation brightness and make the spectral characteristics of the objects more accurately presented.

[0049] Select an area of ​​appropriate size in the original remote sensing image as the training area, crop it, and divide it into a training data set and a validation data set after cropping, for example, select an image of 1000×1000 pixels. Select snow and non-snow samples in the selected image area. In the two categories (snow and non-snow samples), select 70% of the samples for training, and the remaining 30% for validation of the training model. After generating the classified image, use the overall accuracy and global Kappa coefficient to evaluate the validation indicators. The global Kappa coefficient is calculated by the confusion matrix.

[0050] The Normalized Difference Snow Index (NDSI) is the most critical parameter in snow detection. This index has the ability to distinguish ice / snow from clouds. In the visible light range, the reflectivity of ice and snow remains above 50%, appearing white on a single-band grayscale image, while in the near-infrared band, the reflectivity of ice and snow continues to drop to about 20%. At 0.49µm in the blue light band, ice and snow show a reflectivity peak of more than 80%. As the wavelength increases, the reflectivity decreases sharply. This law can be used to construct the Normalized Difference Snow Index (NDSI). The calculation formula of NDSI is: , B 3. B 11They are the 3rd green band and the 11th short-wave infrared band of the spectral data corresponding to the remote sensing image. The threshold used in this application is 0.4, and the threshold of NDSI>0.4 is used to distinguish snow pixels (including snow / ice, snow under shadow, and water bodies) from non-snow pixels.

[0051] Since the threshold value was not determined accurately enough, manual correction was performed based on the normalized snow index method to determine the snow cover area in the training image set.

[0052] The K-Nearest Neighbors (KNN), Random Forest (RF), Support Vector Machine (SVM) and Adaboost were used to establish a machine learning classification model. The training data sets were input into these different initial snow classification models, and the initial snow classification models were trained to obtain the trained snow classification models.

[0053] The implementation principle of the K nearest neighbor classification algorithm is as follows: All samples of known categories are used as references.

[0054] By calculating the distance between the unknown sample and all known samples, the K known samples closest to the unknown sample are selected.

[0055] According to the majority-voting rule, the unknown sample is classified into the category with the largest proportion among the K nearest neighbor samples. K represents the number of nearest neighbor sample instances to be selected. Since the K nearest neighbor classification algorithm only determines the category of the sample to be classified based on the selected K value, rather than relying on the method of distinguishing the class domain to determine the category, the K nearest neighbor algorithm is more suitable for sample sets with overlapping or poor class domains than other methods.

[0056] The implementation principle of the support vector machine algorithm is as follows: The difference between the support vector machine algorithm and the traditional machine learning method is that the support vector machine relies on the principle of structural risk minimization for training. In order to achieve superior generalization performance, the goal of the support vector machine is to strike a balance between the complexity of the model and the learning ability in order to achieve the best processing effect. The solution process of the support vector machine is expressed as a convex quadratic programming problem with linear constraints, and the global optimal solution is calculated based on optimization theory.

[0057] The implementation principle of the random forest algorithm is as follows: The random forest ensemble learning algorithm is built on multiple decision trees. Compared with requiring each decision tree to have too high classification accuracy, the random forest only requires a portion of the trees to have high accuracy. In this algorithm, each decision tree trains and classifies the data independently, and finally all decision trees vote on the data to determine the category it belongs to by majority. During the training process, each decision tree corresponds to a subset of the training data. This technique is called bagging sampling. This method increases the diversity and generalization ability of the model by randomly and repeatedly extracting subsets from the entire training set. When generating the random forest algorithm, each decision tree is trained on a different subset of the data, and the final classification result is obtained by integrating the results.

[0058] The implementation principle of the adaptive enhancement algorithm is as follows: The core concept of Adaptive Boosting (AdaBoost) is to adaptively adjust the sample weights so that new weak classifiers are generated in each iteration, focusing more on the samples that were misclassified in the previous round, thereby continuously improving the overall performance of the model. This process will be repeated and stopped when the preset extremely low error rate is reached or the preset maximum number of iterations is reached, and finally a strong classifier with high accuracy is formed. One of the advantages of AdaBoost is that it can handle high-dimensional data and complex classification problems well, and it is not easy to overfit on this basis. In practical applications, AdaBoost is widely used in binary and multi-classification problems.

[0059] The AdaBoost iterative algorithm process is as follows: Initialize the weight distribution of training data: Assume there are N samples, and initially the weight of each sample is the same, that is, 1 / N.

[0060] Training weak classifiers: In the AdaBoost algorithm, the process of training weak classifiers is very important. In each round, a weak classifier is trained based on the training data with the current weight distribution. In this process, if the sample is correctly classified, its weight will be reduced when constructing the next round of training sets; on the contrary, the weight of the sample that is not correctly classified will increase. Such a dynamic adjustment strategy ensures that subsequent weak classifiers can focus more on the misclassified samples in the previous round, thereby improving the overall performance of the entire model. This iterative process will continue until the preset error rate or the maximum number of iterations is reached, and finally a strong classifier is constructed. This adaptive weight adjustment training method gives AdaBoost excellent performance in solving complex classification problems, while enhancing the stability and generalization ability of the algorithm.

[0061] Construction of strong classifiers: In the process of building strong classifiers, the combination of weak classifiers is not treated equally, but is combined according to their classification error rates. In the final classification decision, weak classifiers with low error rates will be given higher weights, thus affecting the results more significantly; conversely, weak classifiers with high error rates will be given lower weights and have less impact on the final classification. This weight adjustment strategy has demonstrated its effectiveness in practical applications.

[0062] Taking high-resolution optical satellite images of certain areas as original remote sensing images and remote sensing images of another target area as an example, the melting and infiltration of ice and snow are detected. The original remote sensing image has a maximum spatial resolution of 10 meters and has multiple band types including visible light, infrared and SWIR. These bands can distinguish and detect land phenomena such as clouds, ice, water and snow, and monitor these changes. The original remote sensing image is preprocessed. The snow coverage range in the original remote sensing graphics is determined, and images of 1000×1000 pixels are selected. The original remote sensing area is cropped to obtain multiple samples, and the multiple samples are divided into training data sets and validation data sets. Based on the training data sets, the initial snow classification models using the K-Nearest Neighbors algorithm (KNN), the Random Forest algorithm (RF), the Support Vector Machine (SVM) and the Adaptive Boosting algorithm (Adaboost) are trained to obtain multiple different snow classification models.

[0063] Obtain a remote sensing image of the target area, input the remote sensing image into the above-trained different snow classification models, and obtain multiple classification results. The snow classification results obtained using the above snow classification model are as follows: Figure 3 As shown in Figure 1, Figure 2, Figure 3, Figure 4, Figure 5, Figure 6, Figure 7, Figure 8, Figure 9, Figure 10, Figure 11, Figure 12, Figure 13, Figure 14, Figure 15, Figure 16, Figure 17, Figure 18, Figure 19, Figure 20, Figure 21, Figure 22, Figure 23, Figure 24, Figure 25, Figure 26, Figure 27, Figure 28, Figure 29, Figure 29, Figure 30, Figure 31, Figure 32, Figure 33, Figure 34, Figure 35, Figure 36, Figure 37, Figure 38, Figure 39, Figure 39, Figure 40, Figure 41, Figure 42, Figure 43, Figure 44, Figure 45, Figure 46, Figure 47, Figure 48, Figure 49, Figure 49, Figure 50, Figure 51, Figure 52, Figure 53, Figure 54, Figure 55, Figure 56, Figure 57, Figure 58, Figure 59, Figure 59, Figure 51, Figure 5 Figure 4 The fusion result is shown in Figure 5 The snow cover classification results obtained by drone photography of the actual mining area are compared with the fused snow classification results. Figure 6 As shown in Figure a, Figure b are respectively the classification results of the snow cover map obtained by the drone in the actual mining area and the snow classification results after fusion. By comparison, it can be seen that the snow classification accuracy after fusion is higher.

[0064] Table 1 KNN classification confusion matrix

[0065] Table 2 RF classification confusion matrix

[0066] Table 3 Adaboost classification confusion matrix

[0067] Table 4 SVM classification confusion matrix

[0068] Table 5 KAPPA index and overall accuracy

[0069] The climate, soil and other information of the target area were collected to obtain the parameters of the relationship model between meltwater infiltration and snow and ice melting. The obtained model parameters are shown in Table 6. Based on these parameters and the calculated snow cover area, the meltwater infiltration in the target area was calculated.

[0070] Table 6 Model construction parameters

[0071] Furthermore, as a response to the above Figure 1 The embodiment of the present invention provides a device for detecting ice and snow melting and infiltration, such as Figure 7 As shown, the device comprises: The classification module 702 is used to obtain a remote sensing image of the target area, and input the remote sensing image into a plurality of trained snow classification models to obtain a classification result of each snow classification model; A fusion module 704 is used to fuse the classification results of each snow classification model and calculate the snow cover area of ​​the target area according to the fusion results; The calculation module 706 is used to construct a relationship model between meltwater infiltration and ice and snow melting, and calculate the meltwater infiltration in the target area based on the snow cover area of ​​the target area and the relationship model between meltwater infiltration and ice and snow melting.

[0072] The present application provides a device for detecting snow and ice melting infiltration. Compared with the prior art, a plurality of different snow classification models are used to perform snow classification detection on remote sensing images of a target area to obtain a plurality of classification results, which are fused and processed, and the snow cover area is calculated according to the fused results. The calculated snow cover area is input into a relationship model between meltwater infiltration and snow and ice melting to calculate the meltwater infiltration of the target area. Since the results of a plurality of snow classification models are fused, the classification result is more accurate, the obtained snow cover area is more accurate, and the accuracy of meltwater infiltration is improved. At the same time, the relationship model between snow and ice melting and infiltration water volume takes into account the complex conditions of snow and ice melting in open-pit mines in cold regions, and also improves the accuracy of meltwater infiltration.

[0073] In one embodiment, the classification result includes the confidence of multiple category labels corresponding to each pixel in the remote sensing image; the fusion module 704 is also used to: For each snow classification model, obtain the trust corresponding to each category, multiply the confidence of the category label corresponding to each pixel by the trust corresponding to the category label, and obtain the trust distribution data corresponding to each pixel; Based on the trust distribution data corresponding to each pixel point of each snow classification model and the preset conflict factor calculation formula, the conflict factor is calculated; Based on the conflict factor, the trust distribution data corresponding to each pixel of each snow classification model and the preset trust fusion calculation formula, the fused trust distribution data of each pixel is calculated; Based on the fused trust distribution data of each pixel, the comprehensive category of the pixel is determined, and the product of the number of pixels with the comprehensive category of snow and the preset image spatial resolution is taken as the snow cover area of ​​the target area.

[0074] In one embodiment, the device for detecting ice and snow melting and infiltration further comprises: The category trust acquisition module is used to obtain the verification data set corresponding to each snow classification model, input the verification data set into each trained snow classification model, and obtain the model classification results corresponding to each snow classification model; perform positive and negative statistics on the classification results based on the actual classification results of the verification data set and the model classification results, and generate a confusion matrix corresponding to each snow classification model according to the statistical results; calculate based on the confusion matrix corresponding to each snow classification model and the preset trust calculation formula to obtain the trust corresponding to each category of each snow classification model.

[0075] In one embodiment, the calculation module 706 is further configured to: The snow sublimation model was constructed using the surface temperature method; Based on the balance relationship between the total amount of ice and snow, the amount of ice and snow melt, and the surface runoff and the snow sublimation model, a relationship model between meltwater infiltration and ice and snow melt is constructed.

[0076] In one embodiment, the relationship model between meltwater infiltration and ice and snow melting is: , in, R is the meltwater infiltration; is the snow density; S n is the snow area; h is the average thickness of snow; I is the surface runoff, For evaporation, R n is the net radiation, S Energy stored in snow and soil, is the air density, C p is the specific heat of air at constant pressure, f(u x ) is the ground energy transfer equation, ( T s -T a ) is the temperature difference between the snow surface and the air.

[0077] In one embodiment, the device for detecting ice and snow melting and infiltration further comprises: The training module is used to obtain multiple original remote sensing images of the training area and determine the snow-covered area in each original remote sensing image; intercept each original remote sensing image according to a preset image size to obtain multiple training images, and determine the snow classification category and snow-covered area of ​​each training image; based on the multiple training images and their corresponding snow classification categories and snow-covered areas, train multiple different initial snow classification models respectively to obtain multiple trained snow classification models.

[0078] In one embodiment, the multiple snow classification models include snow classification machine learning models constructed using K nearest neighbor algorithm, random forest algorithm, support vector machine algorithm and adaptive boosting algorithm respectively.

[0079] According to one embodiment of the present invention, a storage medium is provided, wherein the storage medium stores at least one executable instruction, and the computer executable instruction can execute the method for detecting ice and snow melting and infiltration in any of the above method embodiments.

[0080] Figure 8 A schematic diagram of the structure of a computer device provided according to an embodiment of the present invention is shown. The specific embodiment of the present invention does not limit the specific implementation of the computer device.

[0081] like Figure 8 As shown, the computer device may include: a processor (processor) 802 , a communication interface (Communications Interface) 804 , a memory (memory) 806 , and a communication bus 808 .

[0082] The processor 802 , the communication interface 804 , and the memory 806 communicate with each other via a communication bus 808 .

[0083] The communication interface 804 is used to communicate with other devices such as clients or other servers.

[0084] The processor 802 is used to execute the program 810, and specifically can execute the relevant steps in the above-mentioned embodiment of the method for detecting ice and snow melting and infiltration.

[0085] Specifically, the program 810 may include program codes, which include computer operation instructions.

[0086] The processor 802 may be a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention. The one or more processors included in the computer device may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0087] The memory 806 is used to store the program 810. The memory 806 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0088] The program 810 may be specifically configured to enable the processor 802 to perform the following operations: Obtain remote sensing images of the target area, and input the remote sensing images into multiple trained snow classification models to obtain classification results of each snow classification model; The classification results of each snow classification model are fused, and the snow cover area of ​​the target area is calculated based on the fusion results; A relationship model between meltwater infiltration and snow and ice melting was constructed, and the meltwater infiltration in the target area was calculated based on the snow cover area of ​​the target area and the relationship model between meltwater infiltration and snow and ice melting.

[0089] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices. In one embodiment, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0090] The above embodiments are only exemplary embodiments of the present application and are not intended to limit the present application. The protection scope of the present application is defined by the claims. Those skilled in the art may make various modifications or equivalent substitutions to the present application within the essence and protection scope of the present application, and such modifications or equivalent substitutions shall also be deemed to fall within the protection scope of the present application.

Claims

1. A method for detecting ice and snow melting and infiltration, characterized in that: include: Acquire a remote sensing image of the target area, and input the remote sensing image into a plurality of different snow classification models that have been trained to obtain a classification result of each snow classification model, wherein the classification result includes the confidence of a plurality of category labels corresponding to each pixel in the remote sensing image; The classification results of each snow classification model are fused, and the snow cover area of ​​the target area is calculated according to the fusion results; Constructing a relationship model between meltwater infiltration and ice and snow melting, and calculating the meltwater infiltration in the target area based on the snow cover area of ​​the target area and the relationship model between meltwater infiltration and ice and snow melting; The step of fusing the classification results of each snow classification model and calculating the snow cover area of ​​the target area according to the fusion results includes: For each of the snow classification models, the confidence corresponding to each category is obtained, and the confidence of the category label corresponding to each pixel is multiplied by the confidence corresponding to the category label to obtain the confidence distribution data corresponding to each pixel; Calculate the conflict factor based on the trust distribution data corresponding to each pixel point of each snow classification model and a preset conflict factor calculation formula; Based on the conflict factor, the trust distribution data corresponding to each pixel of each snow classification model and a preset trust fusion calculation formula, the fused trust distribution data of each pixel is calculated; Based on the fused trust distribution data of each pixel, the comprehensive category of the pixel is determined, and the product of the number of pixels with the comprehensive category of snow and the preset image spatial resolution is used as the snow coverage area of ​​the target area.

2. The method for detecting ice and snow melting and infiltration as claimed in claim 1, characterized in that: Before obtaining the confidence level corresponding to each category for each snow classification model, the method for detecting snow and ice melting infiltration further includes: Obtain a validation data set corresponding to each snow classification model, and input the validation data set into each trained snow classification model to obtain a model classification result corresponding to each snow classification model; Perform positive and negative statistics of the classification results based on the actual classification results of the validation data set and the model classification results, and generate a confusion matrix corresponding to each of the snow classification models according to the statistical results; The confidence corresponding to each category of each snow classification model is obtained by performing calculation based on the confusion matrix corresponding to each snow classification model and a preset confidence calculation formula.

3. The method for detecting ice and snow melting and infiltration according to claim 1, characterized in that: The construction of the relationship model between meltwater infiltration and ice and snow melting includes: The snow sublimation model was constructed using the surface temperature method; Based on the balance relationship between the total amount of ice and snow, the amount of ice and snow melt, and the surface runoff and the snow sublimation model, a relationship model between meltwater infiltration and ice and snow melt is constructed.

4. The method for detecting ice and snow melting and infiltration as claimed in claim 3, characterized in that: The relationship model between the amount of meltwater infiltration and the amount of ice and snow melting is: ， in, R is the meltwater infiltration; is the snow density; S n is the snow area; h is the average thickness of snow; I is the surface runoff, For evaporation, R n is the net radiation, S Energy stored in snow and soil, is the air density, C p is the specific heat of air at constant pressure, f (u x ) is the ground energy transfer equation, ( T s -T a ) is the temperature difference between the snow surface and the air.

5. The method for detecting ice and snow melting and infiltration according to claim 1, characterized in that: Before inputting the remote sensing images into a plurality of trained snow classification models to obtain classification results of each snow classification model, the snow melting and infiltration detection method further includes: Acquire multiple original remote sensing images of the training area and determine the snow covered area in each original remote sensing image; Intercepting each of the original remote sensing images according to a preset image size to obtain a plurality of training images, and determining the snow classification category and snow coverage area of ​​each training image; Based on the multiple training images and their corresponding snow classification categories and snow-covered areas, multiple different initial snow classification models are trained respectively to obtain multiple trained snow classification models.

6. The method for detecting ice and snow melting and infiltration as claimed in claim 5, characterized in that: The multiple snow classification models include snow classification machine learning models constructed using K nearest neighbor algorithm, random forest algorithm, support vector machine algorithm and adaptive boosting algorithm respectively.

7. A device for detecting ice and snow melting and infiltration, characterized in that: include: A classification module, used to obtain a remote sensing image of a target area, and input the remote sensing image into a plurality of trained snow classification models to obtain a classification result of each snow classification model, wherein the classification result includes the confidence of a plurality of category labels corresponding to each pixel in the remote sensing image; A fusion module is used to fuse the classification results of each snow classification model and calculate the snow cover area of ​​the target area according to the fusion results; A calculation module is used to construct a relationship model between meltwater infiltration and ice and snow melting, and calculate the meltwater infiltration of the target area based on the snow cover area of ​​the target area and the relationship model between meltwater infiltration and ice and snow melting; Wherein, the fusion module is also used for: For each of the snow classification models, the confidence corresponding to each category is obtained, and the confidence of the category label corresponding to each pixel is multiplied by the confidence corresponding to the category label to obtain the confidence distribution data corresponding to each pixel; Calculate the conflict factor based on the trust distribution data corresponding to each pixel point of each snow classification model and a preset conflict factor calculation formula; Based on the conflict factor, the trust distribution data corresponding to each pixel of each snow classification model and a preset trust fusion calculation formula, the fused trust distribution data of each pixel is calculated; Based on the fused trust distribution data of each pixel, the comprehensive category of the pixel is determined, and the product of the number of pixels with the comprehensive category of snow and the preset image spatial resolution is used as the snow coverage area of ​​the target area.

8. A storage medium, wherein at least one executable instruction is stored in the storage medium, and the executable instruction enables a processor to execute operations corresponding to the method for detecting ice and snow melting and infiltration as described in any one of claims 1-6.

9. A computer device comprising: A processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the method for detecting ice and snow melting and infiltration as described in any one of claims 1-6.

Images