A method and system for monitoring changes in construction land

Abstract

The present invention discloses a method and system for monitoring changes in construction land, which relates to the field of remote sensing data processing technology. The method and system obtain SAR images and optical images of the area to be monitored; perform multi-phase joint denoising on the SAR image and then perform a first change detection to generate a first difference image of the SAR image; perform a second change detection on the optical image to generate a second difference image of the optical image; and identify the change sub-area in the area to be monitored by combining the first difference image and the second difference image. Compared with traditional methods, the scheme of the present invention is more flexible and has high accuracy and strong rapid response capability. It can adapt to the high-frequency changes in land use at this stage and improve the efficiency of scientific management of cities.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing data processing, and in particular to a method and system for monitoring changes in construction land. Background Art

[0002] With the rapid development of industrialization and urbanization in my country, the area of ​​construction land has increased rapidly. Especially in recent years, the scale of construction land in my country has expanded dramatically, and disorderly expansion has occurred, causing a series of problems such as the decline in land resource utilization, the reduction of cultivated land resources, environmental pollution and ecological damage. Therefore, timely and accurate information on changes in construction land is of great significance for the rational formulation of urban planning programs and the realization of sustainable development of social economy, resource utilization and ecological environment.

[0003] Traditional methods for monitoring changes in construction land mainly include field surveys and satellite image assessment.

[0004] The field investigation method requires staff to go to the site to investigate and collect evidence, and often encounters long distances, difficult roads, and areas that are difficult for personnel to reach. Not only does it put a lot of pressure on time and money costs, but emergencies on the site, such as unnecessary conflicts that affect personal safety, also bring a lot of trouble to the investigation. The disadvantage of the satellite image assessment method is that the satellite remote sensing images used are not real-time, and the images used for comparison are all old images with a long time interval, so the accuracy of assessing illegal land use needs to be improved. Summary of the invention

[0005] The technical problem to be solved by the present invention is how to monitor the changes of construction land quickly and with high precision. In order to solve the above technical problems, the embodiments of the present invention provide a method and system for monitoring changes in construction land, which can adapt to the high-frequency changes in land use at this stage and improve the efficiency of scientific management of cities.

[0006] Specifically, an embodiment of the present invention provides a method for monitoring changes in construction land, including:

[0007] Obtain SAR images and optical images of the area to be monitored;

[0008] Performing a first change detection on the SAR image after performing multi-temporal joint denoising to generate a first difference image of the SAR image;

[0009] Performing a second change detection on the optical image to generate a second difference image of the optical image;

[0010] The first difference image and the second difference image are combined to identify a changed sub-region in the area to be monitored.

[0011] As an improvement of the above solution, the SAR image is subjected to multi-temporal joint denoising and then subjected to first change detection to generate a first difference image of the SAR image, including:

[0012] Performing a first preprocessing on the SAR image to obtain a preprocessed SAR image;

[0013] Performing multi-temporal joint denoising on the preprocessed SAR image to obtain a denoised SAR image;

[0014] A first change detection is performed on the denoised SAR image to generate a first difference image of the SAR image.

[0015] As an improvement of the above solution, the first preprocessing of the SAR image to obtain the preprocessed SAR image includes:

[0016] Performing geometric correction and radiation correction on the SAR image to obtain a corrected SAR image;

[0017] Performing logarithmic transformation on the corrected SAR image to obtain intensity information of the corrected SAR image;

[0018] Subtracting the intensity information in pairs to obtain a registration result;

[0019] According to the registration result, the corrected SAR images are registered pairwise to obtain preprocessed SAR images.

[0020] As an improvement of the above solution, the step of performing multi-temporal joint denoising on the pre-processed SAR image to obtain a denoised SAR image includes:

[0021] Performing Contourlet image transformation and multiplicative noise filtering on the preprocessed SAR image to obtain a preliminary denoised SAR image;

[0022] Extracting temporary ground objects from the preliminary denoised SAR images at different time phases;

[0023] The preliminary denoised SAR image is denoised according to the temporary ground object to obtain a denoised SAR image.

[0024] As an improvement of the above solution, performing a first change detection on the denoised SAR image to generate a first difference image of the SAR image includes:

[0025] Detecting the grayscale value of the denoised SAR image;

[0026] pass Calculate the grayscale difference between different denoised SAR images to obtain the intensity standard deviation; where x ij(t1) is the gray value of the i-th row and j-th column in the denoised SAR image at time t1; x ij (t2) is the grayscale value of the i-th row and j-th column in the denoised SAR image at time t2; time t2 is later than time t1;

[0027] When the grayscale difference is greater than the intensity standard deviation, the corresponding area is marked as a positive change area;

[0028] When the grayscale difference is less than the negative value of the intensity standard deviation, marking the corresponding area as a negative change area;

[0029] A first difference image of the SAR image is generated according to the positive change region and the negative change region.

[0030] As an improvement of the above solution, performing a second change detection on the optical image to generate a second difference image of the optical image includes:

[0031] Performing a second preprocessing on the optical image to obtain a preprocessed optical image;

[0032] Using a radiation normalization method to perform data assimilation on the preprocessed optical image to obtain an assimilated optical image;

[0033] A second change detection is performed on the assimilated optical image to generate a first difference image of the optical image.

[0034] As an improvement of the above solution, the step of performing a second change detection on the assimilated optical image to generate a first difference image of the optical image includes:

[0035] Performing linear transformation on the assimilated optical images at different phases to obtain image difference values ​​of the assimilated optical images;

[0036] Calculating the weight of the invariant feature in the assimilated optical image according to the image difference value;

[0037] When the weight is greater than a preset weight threshold or the number of iterations is not greater than a preset number of iterations threshold, iteratively calculate the image difference value, otherwise stop iterating;

[0038] When the iteration is completed, a first difference image of the optical image is generated according to the image difference value.

[0039] As an improvement of the above solution, the step of combining the first difference image and the second difference image to identify a changed sub-region in the area to be monitored includes:

[0040] Using a pre-trained generative adversarial network to perform heterogeneous remote sensing image change detection on the first difference image and the second difference image to obtain a change detection binary image;

[0041] According to the change detection binary image, a change sub-region of the area to be monitored is identified.

[0042] As an improvement of the above solution, the generative adversarial network includes a generative network and a discriminative network;

[0043] Then, a pre-trained generative adversarial network is used to perform heterogeneous remote sensing image change detection on the first difference image and the second difference image to obtain a change detection binary map, including:

[0044] Performing change detection segmentation on the first difference image and the second difference image through a generative network to obtain a segmented difference image;

[0045] Discriminate the changed pixels and non-changed pixels in the segmented difference image by using a discriminant network;

[0046] A change detection binary image is obtained according to the changed pixels and the non-changed pixels.

[0047] The embodiment of the present invention further provides a system for monitoring changes in construction land, including:

[0048] An image acquisition module is used to acquire SAR images and optical images of the area to be monitored;

[0049] A first difference image generation module, used for performing a first change detection on the SAR image after performing multi-temporal joint denoising to generate a first difference image of the SAR image;

[0050] A second difference image generating module, used for performing a second change detection on the optical image to generate a second difference image of the optical image;

[0051] The changed sub-region monitoring module is used to identify the changed sub-region in the area to be monitored by combining the first difference image and the second difference image.

[0052] Compared with the prior art, the present invention discloses a method and system for monitoring changes in construction land, which obtains SAR images and optical images of the area to be monitored; performs a first change detection on the SAR image after multi-phase joint denoising to generate a first difference image of the SAR image; performs a second change detection on the optical image to generate a second difference image of the optical image; and combines the first difference image and the second difference image to identify the changed sub-area in the area to be monitored. Compared with traditional methods, the scheme of the present invention is more flexible and has high accuracy and strong rapid response capabilities. It can adapt to the high-frequency changes in land use at this stage and improve the efficiency of scientific management of cities. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is a schematic diagram of the steps of a method for monitoring changes in construction land provided by an embodiment of the present invention;

[0054] Figure 2 It is a structural schematic diagram of a construction land change monitoring system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0055] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] In the description of the specification and claims, it is to be understood that the terms first, second, etc. in the specification and claims are only used for the purpose of describing the same technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated, nor necessarily describing the order or time sequence. The terms are interchangeable where appropriate. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features.

[0057] my country launched satellite remote sensing monitoring of illegal construction in all cities across the country for the first time in 2017. Remote sensing monitoring technology is a technology that collects electromagnetic wave information of the environment through aviation or satellites to monitor and identify environmental quality conditions of distant environmental targets. It is an advanced environmental information acquisition technology that is "fast" and "complete" in obtaining large-scale synchronous and dynamic environmental information, which is unmatched and completed by other detection methods. Therefore, it has been increasingly widely used, such as remote sensing monitoring of atmosphere and water quality, investigation of marine oil pollution accidents, investigation of urban thermal environment and thermal pollution in waters, investigation of urban green space, landscape and environmental background, and ecological environment investigation and monitoring.

[0058] Remote sensing monitoring is mainly divided into active remote sensing and passive remote sensing. Some active remote sensing methods use radar sensors to transmit microwave signals to the detection target and receive the backscattered signals after the radar sensors interact with the target to form remote sensing digital images or analog images. According to the characteristics of the device, it can be divided into side-looking radar remote sensing and panoramic radar remote sensing. Side-looking radar remote sensing is widely used, and according to the characteristics of the antenna that transmits microwave beams to the ground, it is divided into real aperture radar system and synthetic aperture radar system. Passive remote sensing is mainly optical remote sensing, which passively receives sunlight reflected from the earth for imaging. It is affected by weather, time and other factors. It cannot be imaged at night and is also greatly disturbed in cloudy and rainy weather.

[0059] Traditional methods for monitoring changes in construction land have at least the following problems: (1) Workers are required to go to the site to investigate and collect evidence, and often encounter areas that are long distances, difficult to pass through, and difficult for personnel to reach. This not only puts a lot of pressure on time and money costs, but also brings a lot of trouble to the investigation due to emergencies on the site, such as unnecessary conflicts that affect personal safety; (2) Complex algorithms require a huge amount of computation, and by using old images separated by a long period of time for monitoring and comparison, the real-time performance is insufficient and the accuracy needs to be improved.

[0060] Based on the above considerations, an embodiment of the present invention provides a method for monitoring changes in construction land. Figure 1 In this embodiment, the construction land change monitoring method is specifically performed through steps S1 to S4:

[0061] S1. Obtain SAR images and optical images of the area to be monitored.

[0062] It should be noted that in the study of change detection, optical remote sensing images have rich spectral information and texture information, and can better retain more detailed information of ground objects. However, optical remote sensing images are greatly affected by factors such as weather and sunlight, and it is difficult to obtain target information in a timely and accurate manner. In microwave remote sensing, synthetic aperture radar (SAR) can overcome the influence of factors such as weather and achieve all-day and all-weather ground observation. However, due to its own imaging characteristics, SAR images are greatly affected by terrain and speckle noise, and the false alarm rate and missed alarm rate are high. By obtaining SAR images and optical images of the area to be monitored, passive optical remote sensing and active microwave remote sensing can be combined to reflect the respective advantages of different types of remote sensing in reflecting surface objects.

[0063] S2. Performing multi-temporal joint denoising on the SAR image and then performing a first change detection to generate a first difference image of the SAR image.

[0064] It should be noted that, for SAR images, the noise is mostly coherent speckle noise, and the characteristics of the noise are mostly multiplicative noise. In the embodiment of the present invention, the multi-temporal SAR images are combined to remove the impact of temporary objects in the image, such as cars on the road, so as to improve the accuracy of the subsequent first change detection.

[0065] It should also be noted that the generation of the first difference image is a key step in detecting the changed area of ​​SAR images. The first change detection can be performed by using pixel-level change detection methods such as image difference method, regression difference method, ratio difference method, vegetation index difference method, tassel transform index difference method, principal component difference method, and change vector method.

[0066] S3. Perform a second change detection on the optical image to generate a second difference image of the optical image.

[0067] S4. Combining the first difference image and the second difference image, identifying a changed sub-region in the area to be monitored.

[0068] In the above scheme, passive optical remote sensing and active microwave remote sensing are combined to identify changes in the monitored area. Compared with traditional methods, it is more flexible and has high accuracy and rapid response capabilities. It can adapt to the current high-frequency changes in land use and improve the efficiency of scientific urban management.

[0069] As an improvement of the above solution, step S2, performing a first change detection on the SAR image after multi-temporal joint denoising to generate a first difference image of the SAR image, is specifically performed through steps S21-S23:

[0070] S21. Perform a first preprocessing on the SAR image to obtain a preprocessed SAR image.

[0071] It should be noted that the SAR image will be affected by the position and attitude of the remote sensing platform, the curvature of the earth's surface, the terrain, atmospheric conditions, etc. during the imaging process, which will cause geometric distortion and radiation distortion of the image. Therefore, it is necessary to perform geometric correction and radiation correction on the SAR image so that subsequent processing can be better performed. In some embodiments, if the source of the SAR image is a geometrically precisely corrected product of satellite data, only radiation correction can be performed on the SAR image.

[0072] S22, performing multi-temporal joint denoising on the pre-processed SAR image to obtain a denoised SAR image.

[0073] S23: Perform a first change detection on the denoised SAR image to generate a first difference image of the SAR image.

[0074] Further, preferably, step S21, performing a first preprocessing on the SAR image to obtain a preprocessed SAR image, comprises:

[0075] Performing geometric correction and radiation correction on the SAR image to obtain a corrected SAR image;

[0076] Performing logarithmic transformation on the corrected SAR image to obtain intensity information of the corrected SAR image;

[0077] Subtracting the intensity information in pairs to obtain a registration result;

[0078] According to the registration result, the corrected SAR images are registered pairwise to obtain preprocessed SAR images.

[0079] For example, by I log-ratio(i, j) = log[I1(i, j)] - log[I2(i, j)] to obtain the registration result; wherein, I1 and I2 are SAR images of the monitored area acquired at different times, I1(i, j) is the amplitude information of the pixel point in the i-th row and j-th column of the SAR image I1, and I2(i, j) is the amplitude information of the pixel point in the i-th row and j-th column of the SAR image I2.

[0080] In the embodiment of the present invention, I1 is used as the reference image and I2 is used as the image to be processed. log-ratio In the paper, the multiplicative speckle noise is effectively suppressed by logarithmically transforming and subtracting two SAR images through the Log-Ratio operator, while the target and background information in the image are retained.

[0081] Preferably, step S22, performing multi-temporal joint denoising on the pre-processed SAR image to obtain a denoised SAR image, comprises:

[0082] Performing Contourlet image transformation and multiplicative noise filtering on the preprocessed SAR image to obtain a preliminary denoised SAR image;

[0083] Extracting temporary ground objects from the preliminary denoised SAR images at different time phases;

[0084] The preliminary denoised SAR image is denoised according to the temporary ground object to obtain a denoised SAR image.

[0085] Exemplarily, the Contourlet transform is represented in the form of inner product. Specifically, the result after the Contourlet transform is in, represents the basis of the Contourlet coefficient space, and x is the input signal.

[0086] According to the embodiment of the present invention, Contourlet transform is a two-dimensional image representation method, which can better capture the line and surface singularities in the two-dimensional image, realize the sparse expression of the two-dimensional image, and can also express smooth curves with fewer coefficients than wavelet transform.

[0087] In some embodiments, after image denoising is completed, a difference image is constructed for multi-phase SAR images according to corresponding characteristics, and by analyzing the difference image, an adaptive threshold method is used to separate the changed and unchanged categories, and the intensity differences of the changed areas are further distinguished.

[0088] Each complex data in the SAR image contains the amplitude and phase information of the surface resolution unit. The roughness, dielectric properties, scatterer structure, etc. of the changed surface will change, thus affecting the amplitude value. Among them, the change in roughness plays a decisive role. The more typical difference change is reflected in the area where buildings are built. Compared with the natural surface with generally low amplitude, the surface roughness and complex dielectric constant of the engineering building with concrete and metal frame are higher, so the backscattering is stronger, which appears as a regular geometric body with high brightness on the SAR image intensity map. Based on the grayscale changes of SAR images in different periods, the difference method can be used to extract the area where the surface properties have changed, so as to compare and analyze the SAR images of different phases in the same region.

[0089] As a preferred implementation, step S23, performing a first change detection on the denoised SAR image to generate a first difference image of the SAR image, includes:

[0090] Detecting the grayscale value of the denoised SAR image;

[0091] pass Calculate the grayscale difference between different denoised SAR images to obtain the intensity standard deviation; where x ij (t1) is the gray value of the i-th row and j-th column in the denoised SAR image at time t1; x ij (t2) is the grayscale value of the i-th row and j-th column in the denoised SAR image at time t2; time t2 is later than time t1;

[0092] When the grayscale difference is greater than the intensity standard deviation, the corresponding area is marked as a positive change area;

[0093] When the grayscale difference is less than the negative value of the intensity standard deviation, marking the corresponding area as a negative change area;

[0094] A first difference image of the SAR image is generated according to the positive change region and the negative change region.

[0095] It should be noted that the difference between the two intensity maps can be obtained by subtracting the grayscale values ​​of the SAR images obtained at different times. In order to avoid misjudgment in the subsequent statistical analysis due to slight differences in intensity information, the threshold is set to the standard deviation of the intensity map difference obtained by statistics. When the difference is greater than the intensity standard deviation, the ground object target in the image is marked, corresponding to the newly added building area; when the difference is less than the negative value of the standard deviation, it generally corresponds to the demolition of the original building or the seasonal changes of vegetation branches and leaves, which can be marked for reference as appropriate. Thus, the intensity difference map of the two SAR images is obtained. Finally, through filtering processing, extraction of change area algorithm, vectorization of spots and geocoding, the dynamic change spots of the surface can be obtained, and the spots suspected of illegal behavior can be checked.

[0096] As a preferred implementation, step S3, performing a second change detection on the optical image to generate a second difference image of the optical image, is specifically performed through steps S31-S33:

[0097] S31, performing a second preprocessing on the optical image to obtain a preprocessed optical image.

[0098] It should be noted that, similar to the SAR images, optical images are also affected by the position and attitude of the remote sensing platform, the curvature of the earth's surface, terrain undulations, atmospheric conditions, etc. during the imaging process.

[0099] Exemplarily, step S31 includes:

[0100] Performing geometric correction and radiation correction on the optical image to obtain a corrected optical image;

[0101] The corrected optical images are registered pairwise to obtain preprocessed optical images.

[0102] S32. Using a radiation normalization method to perform data assimilation on the preprocessed optical image to obtain an assimilated optical image.

[0103] S33, performing a second change detection on the assimilated optical image to generate a first difference image of the optical image.

[0104] Furthermore, step S33, performing a second change detection on the assimilated optical image to generate a first difference image of the optical image, includes:

[0105] Performing linear transformation on the assimilated optical images at different phases to obtain image difference values ​​of the assimilated optical images;

[0106] Calculating the weight of the invariant feature in the assimilated optical image according to the image difference value;

[0107] When the weight is greater than a preset weight threshold or the number of iterations is not greater than a preset number of iterations threshold, iteratively calculate the image difference value, otherwise stop iterating;

[0108] When the iteration is completed, a first difference image of the optical image is generated according to the image difference value.

[0109] For example, by U=α T X and V = β T Y linearly changes the two assimilated optical images X and Y of the monitored area at different phases, and obtains the image difference value of the assimilated optical image through M=UV. Through the canonical correlation analysis theory, the α that meets the condition can be obtained T and β T According to the typical variable properties, it can be known that the M variables are mutually orthogonal, which means that different M variables represent different change information, and the change information decreases in sequence. Therefore, for any pixel point, all the information of its change is Where N represents the total number of bands. Standardize the M variable.

[0110] For unchanged pixels, the value of the M variable should be approximately 0. Since the M variables are independent of each other, they approximately obey the chi-square distribution with N-dimensional freedom, so the chi-square distribution can be used to test whether the pixels have changed. In fact, the above detection method often selects too large a proportion of unchanged areas, so an iterative method is used to calculate M. Each iteration process recalculates the weight of the selected unchanged features, and then updates M until the defined weight ratio is less than a certain threshold, or the number of iterations reaches the given maximum number, and the iteration terminates.

[0111] As a preferred implementation, step S4, identifying a changed sub-region in the area to be monitored by combining the first difference image and the second difference image, includes:

[0112] Using a pre-trained generative adversarial network to perform heterogeneous remote sensing image change detection on the first difference image and the second difference image to obtain a change detection binary image; the generative adversarial network includes a generative network and a discriminative network;

[0113] According to the change detection binary image, a change sub-region of the area to be monitored is identified.

[0114] Furthermore, as a preferred implementation, the generative adversarial network includes a generative network and a discriminative network:

[0115] Then, a pre-trained generative adversarial network is used to perform heterogeneous remote sensing image change detection on the first difference image and the second difference image to obtain a change detection binary map, including:

[0116] Performing change detection segmentation on the first difference image and the second difference image through a generative network to obtain a segmented difference image;

[0117] Discriminate the changed pixels and non-changed pixels in the segmented difference image by using a discriminant network;

[0118] A change detection binary image is obtained according to the changed pixels and the non-changed pixels.

[0119] In the above scheme, for two remote sensing images obtained by different sensors at different times t1 and t2 at the same geographical location, the heterogeneous remote sensing images can be resampled into the same shape after co-registration. The trained encoder is used to extract indifferent features from the two temporally different source images, and the generator is used to construct a reconstructed image with the same number of bands, pixel value range and feature distribution as the target domain image, and the topological structure difference is as small as possible. Heterogeneous remote sensing image change detection is performed based on a generative adversarial network. The network consists of a generative network and a discriminative network. The generative network uses a U-net segmentation network, and the discriminative network uses a convolutional neural network. The segmentation network is used for change detection and segmentation of remote sensing images. The value at the position (I, j, k) on the change detection map represents the pixel at the position (i, k) in the input image, where k = 0 indicates that the detection result is a non-changing pixel, and k = 1 indicates that the detection result is a changing pixel. The change detection map is a binary map, where black represents the area that has not changed, and the white part represents the area that has changed. The discriminant network is used as an auxiliary adversarial network to further constrain the gradient update direction of the segmentation network. Its input is the change detection map from the change detection segmentation network, and its output is a one-dimensional number representing the similarity between the predicted sample and the real sample.

[0120] In some embodiments, based on the intelligent interpretation technology of the remote sensing big data sample library, the heterogeneous remote sensing image change detection is quickly performed to identify the changed sub-areas in the area to be monitored.

[0121] Exemplarily, the intelligent interpretation technology of remote sensing big data sample library includes the following three aspects: data enhancement technology, deep learning semantic segmentation framework, and model training and verification.

[0122] Data augmentation is a method of constructing more samples based on a small number of samples to expand the amount of data. Data augmentation technology can be used to "enhance" a picture into at least four pictures by rotating, flipping, cropping, and other operations on the picture. If similar operations are performed on all pictures in the data set, the data set can be increased to four times the original size.

[0123] In order to solve the problem of information loss caused by continuous downsampling, DeepLab V3+, a classic semantic segmentation method based on dilated convolution, is used. DeepLab V3+ is mainly composed of a backbone network, an atrous spatial pyramid pooling (ASPP) module, and a decoder module. The role of ASPP is to further utilize the multi-scale information of objects. The decoder module refers to the design of UNet, with the purpose of obtaining fine object boundaries. The backbone network of DeepLab V3+ is an improved Xception, and the final output feature map size is one eighth of the original image size.

[0124] The encoding part of the model intends to use the ImageNet pre-training parameters and set ResNet as the encoder. The experimental setting parameters include the number of iterations, optimization function, momentum parameter, weight decay, etc. The polynomial error decay learning strategy can also be used to fine-tune the learning rate, set the initial learning rate, and the learning strategy decay exponent. The image is randomly cropped so that the network input image size is a fixed size, which should match the actual size of the sample. At the same time, data enhancement processing such as horizontal and vertical flipping and rotation is set, and the training batch size, batch regularization method, and the ratio of training samples to test samples are selected.

[0125] More preferably, after the network is trained to convergence, a cropped window prediction method is used to interpret the image of the monitored area, and 9 types of elements are extracted therefrom, including roads, water bodies, grasslands, woodlands, construction sites, structures, greenhouses, cultivated land, and buildings.

[0126] A method for monitoring changes in construction land provided by an embodiment of the present invention identifies changes in the monitored area by combining passive optical remote sensing and active microwave remote sensing. Compared with traditional methods, this method is more flexible and has high accuracy and rapid response capabilities. It can adapt to the current high-frequency changes in land use and improve the efficiency of scientific urban management.

[0127] The embodiment of the present invention also provides a system for monitoring changes in construction land. Figure 2 The construction land change monitoring system includes an image acquisition module 11, a first difference image generation module 12, a second difference image generation module 13 and a change sub-area monitoring module 14, wherein:

[0128] An image acquisition module 11 is used to acquire SAR images and optical images of the area to be monitored;

[0129] A first difference image generation module 12, configured to perform a first change detection on the SAR image after performing multi-temporal joint denoising to generate a first difference image of the SAR image;

[0130] A second difference image generating module 13, configured to perform a second change detection on the optical image to generate a second difference image of the optical image;

[0131] The changed sub-region monitoring module 14 is configured to identify a changed sub-region in the area to be monitored by combining the first difference image and the second difference image.

[0132] As a preferred implementation, the first difference image generating module 12 includes:

[0133] A first preprocessing unit, used for performing a first preprocessing on the SAR image to obtain a preprocessed SAR image;

[0134] A denoising unit, used for performing multi-temporal joint denoising on the pre-processed SAR image to obtain a denoised SAR image;

[0135] The first change detection unit is used to perform a first change detection on the denoised SAR image to generate a first difference image of the SAR image.

[0136] Further, preferably, the first preprocessing unit is specifically used to:

[0137] Performing geometric correction and radiation correction on the SAR image to obtain a corrected SAR image;

[0138] Performing logarithmic transformation on the corrected SAR image to obtain intensity information of the corrected SAR image;

[0139] Subtracting the intensity information in pairs to obtain a registration result;

[0140] According to the registration result, the corrected SAR images are registered pairwise to obtain preprocessed SAR images.

[0141] Preferably, the denoising unit is specifically used for:

[0142] Performing Contourlet image transformation and multiplicative noise filtering on the preprocessed SAR image to obtain a preliminary denoised SAR image;

[0143] Extracting temporary ground objects from the preliminary denoised SAR images at different time phases;

[0144] The preliminary denoised SAR image is denoised according to the temporary ground object to obtain a denoised SAR image.

[0145] As a preferred implementation, the first change detection unit is specifically configured to:

[0146] Detecting the grayscale value of the denoised SAR image;

[0147] pass Calculate the grayscale difference between different denoised SAR images to obtain the intensity standard deviation; where xi j (t1) is the gray value of the i-th row and j-th column in the denoised SAR image at time t1; x ij (t2) is the grayscale value of the i-th row and j-th column in the denoised SAR image at time t2; time t2 is later than time t1;

[0148] When the grayscale difference is greater than the intensity standard deviation, the corresponding area is marked as a positive change area;

[0149] When the grayscale difference is less than the negative value of the intensity standard deviation, marking the corresponding area as a negative change area;

[0150] A first difference image of the SAR image is generated according to the positive change region and the negative change region.

[0151] As a preferred implementation, the second difference image generating module 13 includes:

[0152] A second preprocessing unit, used for performing a second preprocessing on the optical image to obtain a preprocessed optical image;

[0153] A data assimilation unit, used for performing data assimilation on the preprocessed optical image by using a radiation normalization method to obtain an assimilated optical image;

[0154] The second change detection unit is used to perform a second change detection on the assimilated optical image to generate a first difference image of the optical image.

[0155] Further, preferably, the second change detection unit is specifically used to:

[0156] Performing linear transformation on the assimilated optical images at different phases to obtain image difference values ​​of the assimilated optical images;

[0157] Calculating the weight of the invariant feature in the assimilated optical image according to the image difference value;

[0158] When the weight is greater than a preset weight threshold or the number of iterations is not greater than a preset number of iterations threshold, iteratively calculate the image difference value, otherwise stop iterating;

[0159] When the iteration is completed, a first difference image of the optical image is generated according to the image difference value.

[0160] As a preferred implementation, the change sub-area monitoring module 14 includes:

[0161] a change detection binary image generating unit, configured to perform heterogeneous remote sensing image change detection on the first difference image and the second difference image using a pre-trained generative adversarial network to obtain a change detection binary image;

[0162] The changed sub-region identification unit is used to identify the changed sub-region of the area to be monitored according to the change detection binary image.

[0163] Furthermore, the generative adversarial network includes a generative network and a discriminative network; the change detection binary image generating unit is specifically used for:

[0164] Performing change detection segmentation on the first difference image and the second difference image through a generative network to obtain a segmented difference image;

[0165] Discriminate the changed pixels and non-changed pixels in the segmented difference image by using a discriminant network;

[0166] A change detection binary image is obtained according to the changed pixels and the non-changed pixels.

[0167] A construction land change monitoring system provided by an embodiment of the present invention identifies changes in the monitored area by combining passive optical remote sensing and active microwave remote sensing. Compared with traditional methods, it is more flexible and has high accuracy and strong rapid response capabilities. It can adapt to the current high-frequency changes in land use and improve the efficiency of scientific urban management.

[0168] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0169] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for monitoring changes in construction land, characterized in that: include: Obtain SAR images and optical images of the area to be monitored; Performing a first change detection on the SAR image after performing multi-temporal joint denoising to generate a first difference image of the SAR image; Performing a second change detection on the optical image to generate a second difference image of the optical image; The first difference image and the second difference image are combined to identify a changed sub-region in the area to be monitored.

2. A method for monitoring changes in construction land according to claim 1, characterized in that: The step of performing a first change detection after performing multi-temporal joint denoising on the SAR image to generate a first difference image of the SAR image includes: Performing a first preprocessing on the SAR image to obtain a preprocessed SAR image; Performing multi-temporal joint denoising on the preprocessed SAR image to obtain a denoised SAR image; A first change detection is performed on the denoised SAR image to generate a first difference image of the SAR image.

3. A method for monitoring changes in construction land as claimed in claim 2, characterized in that: The performing a first preprocessing on the SAR image to obtain a preprocessed SAR image includes: Performing geometric correction and radiation correction on the SAR image to obtain a corrected SAR image; Performing logarithmic transformation on the corrected SAR image to obtain intensity information of the corrected SAR image; Subtracting the intensity information in pairs to obtain a registration result; According to the registration result, the corrected SAR images are registered pairwise to obtain preprocessed SAR images.

4. A method for monitoring changes in construction land as claimed in claim 2, characterized in that: The performing multi-temporal joint denoising on the pre-processed SAR image to obtain a denoised SAR image includes: Performing Contourlet image transformation and multiplicative noise filtering on the preprocessed SAR image to obtain a preliminary denoised SAR image; Extracting temporary ground objects from the preliminary denoised SAR images at different time phases; The preliminary denoised SAR image is denoised according to the temporary ground object to obtain a denoised SAR image.

5. A method for monitoring changes in construction land as claimed in claim 2, characterized in that: The performing a first change detection on the denoised SAR image to generate a first difference image of the SAR image includes: Detecting the grayscale value of the denoised SAR image; pass Calculate the grayscale difference between different denoised SAR images to obtain the intensity standard deviation; where x ij (t1) is the gray value of the i-th row and j-th column in the denoised SAR image at time t1; x ij (t2) is the grayscale value of the i-th row and j-th column in the denoised SAR image at time t2; time t2 is later than time t1; When the grayscale difference is greater than the intensity standard deviation, the corresponding area is marked as a positive change area; When the grayscale difference is less than the negative value of the intensity standard deviation, marking the corresponding area as a negative change area; A first difference image of the SAR image is generated according to the positive change region and the negative change region.

6. A method for monitoring changes in construction land according to claim 1, characterized in that: The performing a second change detection on the optical image to generate a second difference image of the optical image includes: Performing a second preprocessing on the optical image to obtain a preprocessed optical image; Using a radiation normalization method to perform data assimilation on the preprocessed optical image to obtain an assimilated optical image; A second change detection is performed on the assimilated optical image to generate a first difference image of the optical image.

7. A method for monitoring changes in construction land according to claim 6, characterized in that: The performing a second change detection on the assimilated optical image to generate a first difference image of the optical image includes: Performing linear transformation on the assimilated optical images at different phases to obtain image difference values ​​of the assimilated optical images; Calculating the weight of the invariant feature in the assimilated optical image according to the image difference value; When the weight is greater than a preset weight threshold or the number of iterations is not greater than a preset number of iterations threshold, iteratively calculate the image difference value, otherwise stop iterating; When the iteration is completed, a first difference image of the optical image is generated according to the image difference value.

8. A method for monitoring changes in construction land according to claim 1, characterized in that: The identifying the changed sub-region in the to-be-monitored region by combining the first difference image and the second difference image includes: Using a pre-trained generative adversarial network to perform heterogeneous remote sensing image change detection on the first difference image and the second difference image to obtain a change detection binary image; According to the change detection binary image, a change sub-region of the area to be monitored is identified.

9. A method for monitoring changes in construction land as claimed in claim 8, characterized in that: The generative adversarial network includes a generative network and a discriminative network; Then, a pre-trained generative adversarial network is used to perform heterogeneous remote sensing image change detection on the first difference image and the second difference image to obtain a change detection binary map, including: Performing change detection segmentation on the first difference image and the second difference image through a generative network to obtain a segmented difference image; Discriminate the changed pixels and non-changed pixels in the segmented difference image by using a discriminant network; A change detection binary image is obtained according to the changed pixels and the non-changed pixels.

10. A construction land change monitoring system, characterized in that: include: An image acquisition module is used to acquire SAR images and optical images of the area to be monitored; A first difference image generation module, used for performing a first change detection on the SAR image after performing multi-temporal joint denoising to generate a first difference image of the SAR image; A second difference image generating module, used for performing a second change detection on the optical image to generate a second difference image of the optical image; The changed sub-region monitoring module is used to identify the changed sub-region in the area to be monitored by combining the first difference image and the second difference image.

Images