Regional coverage area calculation method and device, equipment, medium and product

Abstract

The invention discloses a region coverage area calculation method and device, equipment, a medium and a product, and the method comprises the steps: obtaining a to-be-processed remote sensing image; inputting the to-be-processed remote sensing image into an image segmentation model trained by an image training set to obtain segmentation results of various regions of the to-be-processed remote sensing image; according to the segmentation result, calculating the coverage area of each type of region of the to-be-processed remote sensing image; wherein the image training set is obtained by adopting a pre-trained generative adversarial network to expand a historical remote sensing image, and the input of a generator of the generative adversarial network is composed of a random noise vector and feature vectors of various regions of the historical remote sensing image. According to the embodiment of the invention, the key features in the training data can be accurately copied, and the sample quality is improved.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a method, device, equipment, medium and product for calculating regional coverage area. Background Art

[0002] In the process of wireless network deployment and optimization, it is crucial to understand the coverage area of ​​various areas within the cell coverage area. Currently, various area segmentation is achieved by training image segmentation models, and then the coverage area of ​​various areas is calculated. This involves sample data expansion, but traditional sample data expansion methods are difficult to accurately replicate the key features in the training data, resulting in low quality of the expanded samples. Summary of the invention

[0003] The present application provides a method, device, equipment, medium and product for calculating regional coverage area to solve the problem in the prior art that it is difficult to accurately replicate key features in training data, resulting in low quality of expanded samples.

[0004] To achieve the above purpose, the present application provides a method for calculating a regional coverage area, including:

[0005] Acquire remote sensing images to be processed;

[0006] Inputting the remote sensing image to be processed into an image segmentation model trained by an image training set to obtain segmentation results of various regions of the remote sensing image to be processed;

[0007] According to the segmentation result, calculating the coverage area of ​​various regions of the remote sensing image to be processed;

[0008] The image training set is obtained by expanding historical remote sensing images using a pre-trained generative adversarial network, and the input of the generator of the generative adversarial network consists of a random noise vector and feature vectors of various regions of the historical remote sensing images.

[0009] As an improvement of the above scheme, the input of the generator is expressed as:

[0010] b c,i =λ vk b k,i +(1-λ vk )b z,i

[0011] Where b c,i represents the i-th input vector of the generator; λ vk represents the weight of the generator, which is determined by the mean vector of the i-th region; b k,i b represents the feature vector of the i-th region; z,irepresents the i-th random noise vector.

[0012] As an improvement of the above scheme, the loss function of the discriminator of the generative adversarial network is expressed as:

[0013]

[0014] Where, L d represents the loss function of the discriminator, N2 represents the number of real image samples, M2 represents the number of generated image samples, and p r (b x,i ) represents the real image sample b x,i The probability of being identified as a real image, p f (b g,o ) represents the generated image sample b g,i The probability of being classified as a real image.

[0015] As an improvement of the above scheme, the expression of the learning rate of the discriminator of the generative adversarial network is:

[0016]

[0017] In the formula, λ d is the learning rate of the discriminator, L d is the loss function of the discriminator, L g is the loss function of the generator;

[0018] The expression of the learning rate of the generator is:

[0019]

[0020] In the formula, λ g is the learning rate of the generator, L d is the loss function of the discriminator, L g is the loss function of the generator.

[0021] As an improvement of the above solution, the image segmentation model at least includes: an input layer, a convolution layer, a pooling layer, a floating bubble layer and an output layer;

[0022] The input layer is used to preprocess the input image;

[0023] The convolution layer is used to extract features from the image output by the input layer;

[0024] The pooling layer is used to perform feature dimension reduction on the image output by the convolutional layer;

[0025] The floating bubble layer is used to simulate the dynamic behavior of floating bubbles and perform feature processing on the image output by the pooling layer;

[0026] The output layer is used to segment the image output by the floating bubble layer.

[0027] As an improvement of the above solution, the expression of the floating bubble layer is:

[0028]

[0029] In the formula, q f,ij is the characteristic value output by the floating bubble layer, r is the radius of the bubble, φ() is the activation function, and q f,mn is the dynamic weight, q p,i+m,j+n It is the feature value finally output by the pooling layer.

[0030] To achieve the above purpose, the present application also provides a device for calculating a regional coverage area, including:

[0031] An image acquisition module, used for acquiring remote sensing images to be processed;

[0032] A segmentation module, used for inputting the remote sensing image to be processed into an image segmentation model trained by an image training set to obtain segmentation results of various regions of the remote sensing image to be processed;

[0033] A calculation module, used for calculating the coverage area of ​​various regions of the remote sensing image to be processed according to the segmentation result;

[0034] The image training set is obtained by expanding historical remote sensing images using a pre-trained generative adversarial network, and the input of the generator of the generative adversarial network consists of a random noise vector and feature vectors of various regions of the historical remote sensing images.

[0035] To achieve the above-mentioned purpose, an embodiment of the present application also provides a regional coverage area calculation device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and the processor implements the regional coverage area calculation method as described above when executing the computer program.

[0036] To achieve the above-mentioned purpose, an embodiment of the present application also provides a computer-readable storage medium, which includes a stored computer program; wherein, when the computer program is running, it controls the device where the computer-readable storage medium is located to execute the regional coverage area calculation method as described above.

[0037] To achieve the above objectives, an embodiment of the present application further provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the above-mentioned method for calculating the regional coverage area.

[0038] Compared with the prior art, the embodiments of the present application provide a method, device, equipment, medium and product for calculating regional coverage area, which obtains a remote sensing image to be processed; inputs the remote sensing image to be processed into an image segmentation model trained by an image training set to obtain the segmentation results of various regions of the remote sensing image to be processed; and calculates the coverage area of ​​various regions of the remote sensing image to be processed according to the segmentation results; wherein the image training set is obtained by expanding historical remote sensing images using a pre-trained generative adversarial network, and the input of the generator of the generative adversarial network is composed of a random noise vector and feature vectors of various regions of the historical remote sensing image. It can be seen that the embodiments of the present application can accurately replicate the key features in the training data by expanding the image training set by combining a generator composed of feature vectors of various regions of historical remote sensing images, making the images in the image training set more diverse and representative, improving the sample quality, and being able to segment images more accurately, ultimately improving the accuracy of coverage area calculation. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 is a flow chart of a method for calculating regional coverage area provided in an embodiment of the present application;

[0040] Figure 2 It is a structural block diagram of a regional coverage area calculation device provided in an embodiment of the present application;

[0041] Figure 3 It is a structural block diagram of a regional coverage area calculation device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0042] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0043] See also Figure 1 , Figure 1 : is a flow chart of a method for calculating a regional coverage area provided in an embodiment of the present application, wherein the method for calculating a regional coverage area includes:

[0044] S1, obtaining the remote sensing image to be processed;

[0045] S2, inputting the remote sensing image to be processed into an image segmentation model trained by an image training set to obtain segmentation results of various regions of the remote sensing image to be processed;

[0046] S3. Calculating the coverage area of ​​various regions of the remote sensing image to be processed according to the segmentation result;

[0047] The image training set is obtained by expanding historical remote sensing images using a pre-trained generative adversarial network, and the input of the generator of the generative adversarial network consists of a random noise vector and feature vectors of various regions of the historical remote sensing images.

[0048] It is understandable that the remote sensing images to be processed and the historical remote sensing images are high-resolution remote sensing images, which can be obtained through satellite imaging technology, and can ensure coverage of a large area and realize calculations on a large area.

[0049] In an embodiment of the present application, in order to calculate the coverage area of ​​various types of areas in the target area, a remote sensing image to be processed is obtained; the remote sensing image to be processed is input into an image segmentation model trained by an image training set to obtain segmentation results of various types of areas of the remote sensing image to be processed; based on the segmentation results, the coverage area of ​​various types of areas of the remote sensing image to be processed is calculated, which can accurately replicate the key features in the training data, make the images in the image training set more diverse and representative, improve the sample quality, and can more accurately segment the image, ultimately improving the accuracy of coverage area calculation.

[0050] The embodiment of the present application expands the image training set by combining the generator composed of feature vectors of various regions of historical remote sensing images, which can accurately replicate the key features in the training data, making the images in the image training set more diverse and representative, improving the sample quality and applicability, and being able to segment the image more accurately, ultimately improving the accuracy of coverage area calculation. Moreover, the image training set generated by the generative adversarial network is used for the training of the image segmentation model, which improves the utilization efficiency of the data and the generalization ability of the model. In addition, the embodiment of the present application can reduce computing resources and computing time.

[0051] Before training the generative adversarial network, historical remote sensing images are collected and annotated, and the annotated historical remote sensing images are used to train the generative adversarial network. Various areas in the historical remote sensing images are annotated manually, for example, the annotated areas include: residential areas, school areas, commercial areas, other areas, etc.

[0052] Insufficient historical remote sensing images collected can easily lead to poor generalization ability of the model and affect the accuracy of the model. Therefore, the embodiment of the present application adopts a generative adversarial network to expand the historical remote sensing images. Among them, the generative adversarial network includes a generator (G) and a discriminator (D). The generator is responsible for generating new image samples, and the discriminator is used to distinguish between real image samples (i.e., historical remote sensing images collected) and generated image samples (i.e., new image samples generated by the generator). In addition, the input of the generator is composed of a random noise vector and feature vectors of various regions of the historical remote sensing image, which is used to guide the generator to generate diverse and representative images, thereby improving the diversity and representativeness of the images in the image training set.

[0053] For example, the data of historical remote sensing images is represented as: x = {b x,ij |1≤i≤M1,1≤j≤N1}; where b x,ij It represents the spectral information of the pixel in the i-th row and j-th column in the historical remote sensing image, M1 represents the number of rows of the historical remote sensing image, and N1 represents the number of columns of the historical remote sensing image.

[0054] The training steps of the generative adversarial network include:

[0055] S101. Perform cluster analysis on historical remote sensing images to obtain the feature vectors of various regions, denoted as b k = {b k1 ,b k2 ,…,b kn}, where b ki represents the feature vector of the i-th type region, and n represents the dimension of the feature vector.

[0056] Optionally, the cluster centers of each type of region are used as feature vectors, and the feature vectors of each type of region are specifically expressed as:

[0057]

[0058] Where b k,i represents the eigenvector of the i-th region, C i represents the set of all sample points contained in the i-th region, |C i | represents the number of sample points of the i-th category, b x,j represents the jth sample point.

[0059] Furthermore, b x,j The calculation method can be expressed as:

[0060]

[0061] Where b x,j,lrepresents the value of the lth spectral band of the jth sample point, and T represents the number of spectral bands.

[0062] S102, generate a set of random noise vectors, denoted as b z = {b z1 ,b z2 ,…,b zm}, where b 2i represents the i-th random noise vector, and m represents the number of random noise vectors.

[0063] S103, combining the random noise vector with the feature vectors of each type of region to form the input of the generator, the input of the generator is expressed as:

[0064] b c = {b c1 ,b c2 ,…,b cm}

[0065] Optionally, the input of the generator is specifically expressed as:

[0066] b c,i =λ vk b k,i +(1-λ vk )b z,i

[0067] Where b c,i represents the i-th input vector of the generator; λ vk represents the weight of the generator, which is determined by the mean vector of the i-th region; b k,i b represents the feature vector of the i-th region; z,i represents the i-th random noise vector.

[0068] Furthermore, the calculation method of the generator weight λ can be expressed as:

[0069]

[0070] Where b u,i Indicates the parameters calculated based on the input data, the calculation method is:

[0071]

[0072] In the formula, μ i represents the mean vector of the i-th region, b x,j represents the jth sample point, C i represents the set of all sample points contained in the i-th region, |C i | represents the number of sample points of the i-th category.

[0073] S104, the generator receives an input consisting of a random noise vector and feature vectors of various regions of the historical remote sensing image, and generates a set of new remote sensing image samples, which are expressed as:

[0074] b g = {b g1 ,b g2 ,…,b gm}

[0075] That is, the output of the generator is:

[0076] b g,i =G(b c,i θ g )

[0077] Where b g,i represents the i-th generated image sample, G(;θ g ) represents the function of the generator, θ g Represents the parameters of the generator.

[0078] S105, mixing the generated image samples with the real image samples and inputting them into the discriminator; wherein the discriminator outputs the probability that each image sample is a real image, which can be expressed as:

[0079] p r (b x,i )=D(b x,i θ d )

[0080] p f (b g,i )=D(b g,i θ d )

[0081] In the formula, p r (b x,i ) represents the real image sample b x,i The probability of being identified as a real image, p f (b g,i ) represents the generated image sample b g,i The probability of being judged as a real image, D(;θ d ) represents the function of the discriminator, θ d Represents the parameters of the discriminator.

[0082] Furthermore, the loss function of the discriminator is expressed as:

[0083]

[0084] Where, L d represents the loss function of the discriminator, N2 represents the number of real image samples, M2 represents the number of generated image samples, and pr (b x,o ) represents the real image sample b x,i The probability of being judged as a real image, p f (b g,o ) represents the generated image sample b g,i The probability of being classified as a real image.

[0085] Furthermore, the calculation method of the discriminator function D() can be expressed as:

[0086] D(b x,i θ d )=Sig(w·φ(b x,i )+b)

[0087] In the formula, Sig() represents the Sigmoid activation function, w represents the weight vector of the discriminator, φ() represents the feature extraction function, and b represents the bias term.

[0088] Furthermore, the calculation method of the feature extraction function φ() can be expressed as:

[0089] φ(b x,i )=ReLU(W1b x,i +b1)

[0090] Where ReLU() represents the ReLU activation function, W1 represents the weight matrix of the generator, and b1 represents the first bias vector.

[0091] S106, the generator adjusts its own parameters according to the feedback from the discriminator; the goal is to make the discriminator unable to distinguish between the generated image samples and the real image samples;

[0092] The calculation method of the generator's loss function can be expressed as:

[0093]

[0094] Where, L g represents the loss function of the generator, M2 represents the number of generated image samples, and p f (b g,i ) represents the generated image sample b g,i The probability of being classified as a real image.

[0095] Furthermore, the output b of the generator g,i The calculation method is:

[0096] b g,i =G(b c,i θ g )

[0097] Moreover, the calculation method of the generator function G() can be expressed as:

[0098] G(b c,i θ g )=tanh(W2·ReLU(W1b c,i +b1)+b2)

[0099] In the formula, tanh() represents the hyperbolic tangent activation function, W2 represents the second weight matrix, b2 represents the second bias vector, ReLU is the activation function, W1 represents the weight matrix of the generator, and b1 represents the first bias vector.

[0100] S107, in each process of adjusting the parameters of the generator and the discriminator, adaptively adjust the learning rate λ of the discriminator according to the current training results d and the learning rate λ of the generator g , which can be expressed as:

[0101]

[0102] In the formula, λ d represents the learning rate of the discriminator, λ g Represents the learning rate of the generator. The learning rate of the discriminator and the learning rate of the generator are used in the training process of the discriminator and the generator, and are set when the parameters are adjusted.

[0103] S108, repeat the above steps until the preset first stop iteration condition is met, indicating that the generative adversarial network training is completed. In one embodiment, the preset first stop iteration condition is reaching a preset first maximum number of iterations, preferably, the preset first maximum number of iterations is set to 1000 times.

[0104] After the historical remote sensing images are expanded by using the generative adversarial network to obtain an image training set, the images in the image training set are used to train the image segmentation model. The embodiment of the present application adopts an image segmentation model based on a floating bubble morphology algorithm to achieve high-precision remote sensing image segmentation. The image segmentation model based on the floating bubble morphology algorithm includes an input layer, a convolution layer, a pooling layer, a floating bubble layer and an output layer;

[0105] The input layer is used to preprocess the input image;

[0106] The convolution layer is used to extract features from the image output by the input layer;

[0107] The pooling layer is used to perform feature dimension reduction on the image output by the convolutional layer;

[0108] The floating bubble layer is used to simulate the dynamic behavior of floating bubbles and perform feature processing on the image output by the pooling layer;

[0109] The output layer is used to segment the image output by the floating bubble layer.

[0110] The embodiment of the present application uses a floating bubble layer to simulate the dynamic behavior of bubbles, thereby enhancing the expression ability of image features, improving the accuracy and efficiency of segmentation, and can effectively process complex or subtle image features, especially in complex environments, which may cause the segmentation results to be inaccurate.

[0111] Specifically, the training steps of the image segmentation model include:

[0112] S201, input the image in the image training set into the input layer for preprocessing; let the input image be q x , which can be expressed as:

[0113] q x = {q x,ij ∣1≤i≤M3,1≤j≤N3}

[0114] In the formula, q x,ij represents the spectral information of the pixel in the i-th row and j-th column, M3 represents the number of rows of the image, and N3 represents the number of columns of the image.

[0115] Furthermore, the preprocessing includes normalizing the input image so that the pixel values ​​are between [0, 1], which can be expressed as:

[0116]

[0117] In the formula, q ′ x,ij is the normalized pixel value, min(q x ) and max(q x ) are the minimum and maximum values ​​of the input image respectively.

[0118] S202: Input the image output by the input layer into the convolution layer for feature extraction. The convolution layer includes several convolution layers. Exemplary:

[0119] In the convolution layer, the first convolution layer extracts preliminary features through convolution operation, assuming the convolution kernel is q w1 , the convolution operation can be expressed as:

[0120]

[0121] In the formula, q c1,ij is the eigenvalue of the first convolutional layer output, Re() is the ReLU activation function, k is the size of the convolution kernel, q b1 is the first bias term.

[0122] Furthermore, the convolution kernel q w1,mn The calculation method can be expressed as:

[0123]

[0124] In the formula, is the input convolution kernel q w1,mn The dimension of the feature, N(0,1) represents a normal distribution with a mean of 0 and a variance of 1.

[0125] Furthermore, the second convolutional layer further extracts higher-level features, assuming that the convolution kernel is q w2 , the convolution operation can be expressed as:

[0126]

[0127] In the formula, q c2,ij is the feature value output by the second convolutional layer, q b2 is the second bias term.

[0128] Similarly, the convolution kernel q w2,mn The calculation method can be expressed as:

[0129]

[0130] In the formula, is the input convolution kernel q w2,mn The dimension of the feature, N(0,1) represents a normal distribution with a mean of 0 and a variance of 1.

[0131] Furthermore, by analogy, multiple convolutional layers are superimposed, so that the image segmentation model can gradually capture the complex features of the image.

[0132] S203, input the image output by the convolution layer into the corresponding pooling layer for dimensionality reduction. Exemplary:

[0133] Set the pooling layer dimensionality reduction operation. The first pooling layer reduces the dimension of the features of the first convolutional layer and retains important information. Set the pooling window size to p. The pooling operation can be expressed as:

[0134]

[0135] In the formula, q p1,ij It is the feature value output by the first pooling layer.

[0136] Furthermore, the selection criteria for the pooling window p are:

[0137]

[0138] Where H and W are the height and width of the input feature map, respectively.

[0139] Furthermore, the second pooling layer reduces the dimension of the features of the second convolutional layer, further reducing the amount of calculation, which can be expressed as:

[0140]

[0141] In the formula, q p2,ij It is the feature value output by the second pooling layer.

[0142] Further, by analogy, multiple pooling layers are stacked to improve computational efficiency while retaining important features.

[0143] S204: Input the image output by the last pooling layer into the floating bubble layer for feature processing. The floating bubble layer simulates the dynamic behavior of bubbles and sets the dynamic weight of the floating bubble layer as q f , the expression of the floating bubble layer is:

[0144]

[0145] In the formula, q f,ij is the characteristic value output by the floating bubble layer, r is the radius of the bubble, φ() is the activation function, and q f,mn is the dynamic weight, q p,i+m,j+n is the feature value finally output by the pooling layer. Here, there are two pooling layers, so q p,i+m,j+n =q p2,i+m,j+n .

[0146] Furthermore, the dynamic weight q f,mn The calculation method can be expressed as:

[0147]

[0148] In the formula, α mn Represents dynamic weight parameters, which are learned through training.

[0149] S205: Input the image output by the floating bubble layer into the output layer to generate a segmented image. Suppose the weight of the output layer is q w3 , the output calculation can be expressed as:

[0150]

[0151] In the formula, q y,ij The segmentation result generated by the output layer, q b3 is the third bias term, and σ() is the activation function.

[0152] S206, using an adaptive fitness function to dynamically adjust network parameters according to the segmentation results, the fitness function L is calculated as follows:

[0153]

[0154] In the formula, q t,ij is the true segmentation label.

[0155] Furthermore, in each iteration, the fitness function is calculated, and the parameters of the image segmentation model are adjusted according to the feedback of the fitness function. The parameter adjustment method can be expressed as:

[0156]

[0157] In the formula, q w,mn is the parameter before adjustment, q w,mn ' is the adjusted parameter, α bk is the learning rate of the image segmentation model. Preferably, α bk Set to 0.03.

[0158] S207, repeat the above steps until the preset second stop iteration condition is met, indicating that the image segmentation model training is completed. In one embodiment, the preset second stop iteration condition is reaching a preset second maximum number of iterations, preferably, the preset second maximum number of iterations is set to 1000 times.

[0159] In an optional embodiment, the calculating, according to the segmentation result, the coverage area of ​​each type of region of the remote sensing image to be processed includes:

[0160] Performing pixel-level statistics on the segmentation results to obtain the number of pixels in various regions of the remote sensing image to be processed;

[0161] The coverage area of ​​each type of region of the remote sensing image to be processed is calculated according to the number of pixels and the preset pixel resolution.

[0162] Specifically, the number of pixels in each area of ​​the remote sensing image to be processed is calculated according to the following formula:

[0163]

[0164] Where N e is the number of pixels in the area of ​​category c, δ is the indicator function, when q y,ij =c, the value is 1, otherwise it is 0.

[0165] The calculation method of the indicator function δ can be expressed as:

[0166]

[0167] According to the following formula, the coverage area of ​​various regions of the remote sensing image to be processed is calculated:

[0168] A c =N c ·R2

[0169] In the formula, A c is the coverage area of ​​the region of category c, and R is the pixel resolution.

[0170] A method for calculating regional coverage area provided in an embodiment of the present application comprises the following steps: obtaining a remote sensing image to be processed; inputting the remote sensing image to be processed into an image segmentation model trained by an image training set to obtain segmentation results of various regions of the remote sensing image to be processed; and calculating the coverage area of ​​various regions of the remote sensing image to be processed based on the segmentation results; wherein the image training set is obtained by expanding historical remote sensing images using a pre-trained generative adversarial network, and the input of the generator of the generative adversarial network is composed of a random noise vector and feature vectors of various regions of the historical remote sensing image. It can be seen that the embodiment of the present application can accurately replicate the key features in the training data by expanding the image training set by combining a generator composed of feature vectors of various regions of historical remote sensing images, making the images in the image training set more diverse and representative, improving the sample quality, and being able to segment images more accurately, thereby ultimately improving the accuracy of coverage area calculation.

[0171] See also Figure 2 , Figure 2 : is a structural block diagram of a regional coverage area calculation device 10 provided in an embodiment of the present application, and the regional coverage area calculation device 10 includes:

[0172] An image acquisition module 11 is used to acquire remote sensing images to be processed;

[0173] The segmentation module 12 is used to input the remote sensing image to be processed into an image segmentation model trained by an image training set to obtain segmentation results of various regions of the remote sensing image to be processed;

[0174] A calculation module 13, used for calculating the coverage area of ​​various regions of the remote sensing image to be processed according to the segmentation result;

[0175] The image training set is obtained by expanding historical remote sensing images using a pre-trained generative adversarial network, and the input of the generator of the generative adversarial network consists of a random noise vector and feature vectors of various regions of the historical remote sensing images.

[0176] Optionally, the input of the generator is expressed as:

[0177] b c,i =λ vk b k,i +(1-λ vk )b z,i

[0178] Where b c,i represents the i-th input vector of the generator; λ vk represents the weight of the generator, which is determined by the mean vector of the i-th region; b k,i b represents the feature vector of the i-th region; z,i represents the i-th random noise vector.

[0179] Optionally, the loss function of the discriminator of the generative adversarial network is expressed as:

[0180]

[0181] Where, L d represents the loss function of the discriminator, N2 represents the number of real image samples, M2 represents the number of generated image samples, and p r (b x,o ) represents the real image sample b x,i The probability of being judged as a real image, p f (b g,o ) represents the generated image sample b g,i The probability of being classified as a real image.

[0182] Optionally, the expression of the learning rate of the discriminator of the generative adversarial network is:

[0183]

[0184] In the formula, λ d is the learning rate of the discriminator, L d is the loss function of the discriminator, L g is the loss function of the generator;

[0185] The expression of the learning rate of the generator is:

[0186]

[0187] In the formula, λ g is the learning rate of the generator, L d is the loss function of the discriminator, L g is the loss function of the generator.

[0188] Optionally, the image segmentation model includes at least: an input layer, a convolution layer, a pooling layer, a floating bubble layer and an output layer;

[0189] The input layer is used to preprocess the input image;

[0190] The convolution layer is used to extract features from the image output by the input layer;

[0191] The pooling layer is used to perform feature dimension reduction on the image output by the convolutional layer;

[0192] The floating bubble layer is used to simulate the dynamic behavior of floating bubbles and perform feature processing on the image output by the pooling layer;

[0193] The output layer is used to segment the image output by the floating bubble layer.

[0194] Optionally, the expression of the floating bubble layer is:

[0195]

[0196] In the formula, q f,ij is the characteristic value output by the floating bubble layer, r is the radius of the bubble, φ() is the activation function, and q f,mn is the dynamic weight, q p,i+m,j+n It is the feature value finally output by the pooling layer.

[0197] It is worth noting that the working process of each module in the regional coverage area calculation device 10 described in the embodiment of the present application can refer to the working process of the regional coverage area calculation method described in the above embodiment, and will not be repeated here.

[0198] The embodiment of the present application provides a device 10 for calculating regional coverage area, which obtains a remote sensing image to be processed; inputs the remote sensing image to be processed into an image segmentation model trained by an image training set to obtain segmentation results of various regions of the remote sensing image to be processed; and calculates the coverage area of ​​various regions of the remote sensing image to be processed according to the segmentation results; wherein the image training set is obtained by expanding historical remote sensing images using a pre-trained generative adversarial network, and the input of the generator of the generative adversarial network is composed of a random noise vector and feature vectors of various regions of the historical remote sensing image. It can be seen that the embodiment of the present application expands the image training set by combining a generator composed of feature vectors of various regions of historical remote sensing images, which can accurately replicate the key features in the training data, making the images in the image training set more diverse and representative, improving the sample quality, and being able to segment images more accurately, ultimately improving the accuracy of coverage area calculation.

[0199] In addition, an embodiment of the present application further provides a computer-readable storage medium, which includes a stored computer program; wherein, when the computer program is running, it controls the device where the computer-readable storage medium is located to execute the regional coverage area calculation method as described in any of the above embodiments.

[0200] In addition, an embodiment of the present application further provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the method for calculating the regional coverage area as described in any of the above embodiments.

[0201] See also Figure 3 , Figure 3 : is a structural block diagram of a regional coverage area calculation device 20 provided in an embodiment of the present application, and the regional coverage area calculation device 20 includes: a processor 21, a memory 22, and a computer program stored in the memory 22 and executable on the processor 21. When the processor 21 executes the computer program, the steps in the above-mentioned regional coverage area calculation method embodiment are implemented. Alternatively, when the processor 21 executes the computer program, the functions of each module / unit in the above-mentioned device embodiments are implemented.

[0202] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory 22 and executed by the processor 21 to complete the present application. The one or more modules / units may be a series of computer program instruction segments capable of completing specific functions, which are used to describe the execution process of the computer program in the area coverage area calculation device 20.

[0203] The regional coverage area calculation device 20 may include, but is not limited to, a processor 21 and a memory 22. Those skilled in the art will appreciate that the schematic diagram is merely an example of the regional coverage area calculation device 20 and does not constitute a limitation on the regional coverage area calculation device 20. The regional coverage area calculation device 20 may include more or fewer components than shown in the diagram, or may combine certain components, or different components. For example, the regional coverage area calculation device 20 may also include input and output devices, network access devices, buses, and the like.

[0204] The processor 21 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor 21 is the control center of the regional coverage area calculation device 20, and uses various interfaces and lines to connect various parts of the entire regional coverage area calculation device 20.

[0205] The memory 22 can be used to store the computer program and / or module. The processor 21 implements various functions of the regional coverage area calculation device 20 by running or executing the computer program and / or module stored in the memory 22 and calling the data stored in the memory 22. The memory 22 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory 22 can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0206] Wherein, if the module / unit integrated in the regional coverage area calculation device 20 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor 21. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal and software distribution medium, etc.

[0207] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present application, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art can understand and implement it without paying any creative work.

[0208] The above is a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications are also considered to be within the scope of protection of the present application.

Claims

1. A method for calculating regional coverage area, characterized in that: include: Acquire remote sensing images to be processed; Inputting the remote sensing image to be processed into an image segmentation model trained by an image training set to obtain segmentation results of various regions of the remote sensing image to be processed; According to the segmentation result, calculating the coverage area of ​​various regions of the remote sensing image to be processed; The image training set is obtained by expanding historical remote sensing images using a pre-trained generative adversarial network, and the input of the generator of the generative adversarial network consists of a random noise vector and feature vectors of various regions of the historical remote sensing images.

2. The method for calculating the regional coverage area according to claim 1, characterized in that: The input of the generator is expressed as: b c,i =λ vk b k,i +(1-l vk )b z,i Where b c,i represents the i-th input vector of the generator; λ vk represents the weight of the generator, which is determined by the mean vector of the i-th region; b k,i b represents the feature vector of the i-th region; z,i represents the i-th random noise vector.

3. The method for calculating the regional coverage area according to claim 1, characterized in that: The loss function of the discriminator of the generative adversarial network is expressed as: Where, L d represents the loss function of the discriminator, N2 represents the number of real image samples, M2 represents the number of generated image samples, and p r (b x,i ) represents the real image sample b x,i The probability of being identified as a real image, p f (b g,i ) represents the generated image sample b g,i The probability of being classified as a real image.

4. The method for calculating the regional coverage area according to claim 1, characterized in that: The expression of the learning rate of the discriminator of the generative adversarial network is: In the formula, λ d is the learning rate of the discriminator, L d is the loss function of the discriminator, L g is the loss function of the generator; The expression of the learning rate of the generator is: In the formula, λ g is the learning rate of the generator, L d is the loss function of the discriminator, L g is the loss function of the generator.

5. The method for calculating the regional coverage area according to claim 1, characterized in that: The image segmentation model at least includes: an input layer, a convolution layer, a pooling layer, a floating bubble layer and an output layer; The input layer is used to preprocess the input image; The convolution layer is used to extract features from the image output by the input layer; The pooling layer is used to perform feature dimension reduction on the image output by the convolutional layer; The floating bubble layer is used to simulate the dynamic behavior of floating bubbles and perform feature processing on the image output by the pooling layer; The output layer is used to segment the image output by the floating bubble layer.

6. The method for calculating the regional coverage area according to claim 5, characterized in that: The expression of the floating bubble layer is: In the formula, q f,ij is the characteristic value output by the floating bubble layer, r is the radius of the bubble, φ() is the activation function, and q f,mn is the dynamic weight, q p,i+m,j+n It is the feature value finally output by the pooling layer.

7. A device for calculating regional coverage area, characterized in that: include: An image acquisition module, used for acquiring remote sensing images to be processed; A segmentation module, used for inputting the remote sensing image to be processed into an image segmentation model trained by an image training set to obtain segmentation results of various regions of the remote sensing image to be processed; A calculation module, used for calculating the coverage area of ​​various regions of the remote sensing image to be processed according to the segmentation result; The image training set is obtained by expanding historical remote sensing images using a pre-trained generative adversarial network, and the input of the generator of the generative adversarial network consists of a random noise vector and feature vectors of various regions of the historical remote sensing images.

8. A device for calculating area coverage, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the method for calculating the regional coverage area according to any one of claims 1 to 6 when executing the computer program.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program; wherein, when the computer program is run, it controls the device where the computer-readable storage medium is located to execute the method for calculating the regional coverage area according to any one of claims 1 to 6.

10. A computer program product, characterized in that The method comprises a computer program / instruction, which, when executed by a processor, implements the method for calculating the regional coverage area as claimed in any one of claims 1 to 6.

Images