SAR image quality evaluation method and device and storage medium

Abstract

The invention discloses an SAR image quality evaluation method and device and a storage medium. The SAR image quality evaluation method comprises the steps of collecting an SAR image corresponding to a target area by using an SAR satellite; the SAR image is input to a pre-trained quality evaluation model, a plurality of first feature maps corresponding to the SAR image are determined by using each time convolution model in the quality evaluation model, the time convolution model comprises multiple layers of time convolution networks, and the receptive fields of the layers of time convolution networks are different; and performing quality evaluation on the SAR image corresponding to the target area based on the plurality of first feature maps by using a quality evaluation model. The SAR image is subjected to feature extraction by using a plurality of time convolution models, so that the accuracy of feature extraction can be ensured to the greatest extent.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of SAR image processing, in particular to a quality evaluation method and device of a SAR image and a storage medium. BACKGROUND

[0002] A SAR image is a high-resolution radar image obtained through active microwave remote sensing technology. Unlike optical sensors, a SAR satellite does not rely on sunlight, but images by emitting electromagnetic waves and receiving backscattering signals of a target area. Since a SAR image is not dependent on optical imaging, but is imaged by emitting electromagnetic waves, a SAR satellite can work all day and is not limited by lighting conditions. In addition, the microwave signals emitted by a SAR satellite can penetrate clouds and smoke, and are suitable for most areas.

[0003] Among them, the abscissa of the SAR image usually represents the azimuth direction, and the ordinate represents the slant range, and the pixel value of each pixel in the SAR image reflects the backscattering intensity of the corresponding position target. And where the azimuth direction of the SAR image refers to the direction of the flight trajectory of the SAR satellite, which is directly related to the time sequence. Thus, the essence of the spatial information of the SAR image is converted from the time sequence signal.

[0004] Further, in the quality evaluation of the SAR image at present, the SAR image is usually processed by using a convolutional neural network (CNN) or Vit. Although the convolutional neural network (CNN) and Vit perform well in processing spatial distribution data (for example, optical remote sensing images), there are certain limitations in processing SAR images whose essence is time sequence signals. For example, since the pixel value of an optical remote sensing image directly reflects the spatial radiation characteristics of a ground object, the convolutional neural network (CNN) can effectively extract the features of adjacent pixels belonging to the same ground object, or the self-attention mechanism of ViT can effectively extract the ground object distribution features across the entire image.

[0005] However, since the essence of the spatial information of the SAR image is converted from the time sequence signal, directly processing the SAR image by using the convolutional neural network (CNN) or Vit and using it to evaluate the quality of the SAR image may have inaccuracy problems.

[0006] A method and system for constructing a SAR image quality evaluation model are disclosed in CN118967632A. The method includes: constructing a training set and a test set; constructing a machine learning model, based on the training set, using an improved Harris eagle optimization algorithm to iteratively optimize the hyperparameters of the machine learning model to obtain an optimal hyperparameter combination; based on the optimal hyperparameter combination and the test set, iteratively optimizing the weight coefficients of the machine learning model to obtain an optimal weight coefficient combination; and based on the optimal hyperparameter combination and the optimal weight coefficient combination, obtaining an evaluation model.

[0007] A SAR image quality evaluation method based on correlation and discriminability is disclosed in CN117291873A. The method includes the following steps: S1, obtaining two single-view complex images of the same ground object observed by SAR at two different times, one without noise disturbance from a same-frequency civil radiation source or active interference source and the other with noise disturbance, calculating the correlation coefficient of the two registered images, and obtaining the signal-to-noise ratio corresponding to the correlation coefficient by establishing a quantitative relationship between the correlation coefficient and the signal-to-noise ratio; S2, establishing the form of the signal-to-noise ratio and the signal-to-clutter ratio, and converting the signal-to-noise ratio into the signal-to-clutter ratio of the SAR image disturbed by noise; S3, calculating the national image interpretability rating scale based on the SAR image according to the signal-to-clutter ratio, converting it into the NIIRS of the corresponding infrared detector, and then calculating the target discrimination probability for the target to be discriminated in the image.

[0008] The existing technology has the technical problem that the spatial information of a SAR image is essentially converted from a time-series signal, so if a model suitable for processing spatial analysis data is used to evaluate the quality of a SAR image, the evaluation may not be accurate. SUMMARY

[0009] Embodiments of the present disclosure provide a SAR image quality evaluation method, device and storage medium to at least solve the technical problem that the spatial information of a SAR image is essentially converted from a time-series signal, so if a model suitable for processing spatial analysis data is used to evaluate the quality of a SAR image, the evaluation may not be accurate.

[0010] According to an aspect of embodiments of the present disclosure, a method for quality assessment of a SAR image is provided. The method includes: collecting, by a SAR satellite, a SAR image corresponding to a target region, wherein an abscissa of the SAR image represents a time sequence related azimuth, an ordinate of the SAR image represents a slant range, and a pixel value of the SAR image represents a backscattering intensity corresponding to the azimuth and the slant range; inputting the SAR image into a pre-trained quality assessment model, and determining a plurality of first feature maps corresponding to the SAR image by using each time convolution model in the quality assessment model, wherein the time convolution model includes a plurality of time convolution networks, and a receptive field of each time convolution network is different; and performing quality assessment on the SAR image corresponding to the target region based on the plurality of first feature maps and by using the quality assessment model.

[0011] According to another aspect of embodiments of the present disclosure, a storage medium is also provided. The storage medium includes a stored program, wherein the program is executed by a processor when the program is running.

[0012] According to another aspect of embodiments of the present disclosure, a device for quality assessment of a SAR image is also provided. The device includes: a SAR image collecting module configured to collect, by a SAR satellite, a SAR image corresponding to a target region, wherein an abscissa of the SAR image represents a time sequence related azimuth, an ordinate of the SAR image represents a slant range, and a pixel value of the SAR image represents a backscattering intensity corresponding to the azimuth and the slant range; a first feature map determining module configured to input the SAR image into a pre-trained quality assessment model, and determine a plurality of first feature maps corresponding to the SAR image by using each time convolution model in the quality assessment model, wherein the time convolution model includes a plurality of time convolution networks, and a receptive field of each time convolution network is different; and a quality assessment module configured to perform quality assessment on the SAR image corresponding to the target region based on the plurality of first feature maps and by using the quality assessment model.

[0013] According to another aspect of the embodiments of the present disclosure, a device for quality assessment of a SAR image is also provided, including: a processor; and a memory connected with the processor, configured to provide the processor with instructions for processing the following processing steps: collecting, by a SAR satellite, a SAR image corresponding to a target region, wherein the abscissa of the SAR image represents a time sequence related azimuth, the ordinate of the SAR image represents a slant range, and the pixel value of the SAR image represents a backscattering intensity corresponding to the azimuth and the slant range; inputting the SAR image into a pre-trained quality assessment model, and determining a plurality of first feature maps corresponding to the SAR image by using each time convolution model in the quality assessment model, wherein the time convolution model includes a plurality of time convolution networks, and the receptive field of each layer of the time convolution network is different; and performing quality assessment on the SAR image corresponding to the target region based on the plurality of first feature maps and by using the quality assessment model.

[0014] The present application discloses a method for quality assessment of a SAR image. First, a SAR satellite collects a SAR image corresponding to a target region. Then, the SAR satellite inputs the collected SAR image into a pre-trained quality assessment model, and determines a plurality of first feature maps corresponding to the SAR image by using each time convolution model in the quality assessment model. Finally, the SAR satellite performs quality assessment on the SAR image corresponding to the target region based on the plurality of first feature maps and by using the quality assessment model.

[0015] The present application discloses a quality assessment model, and the quality assessment model includes a plurality of time convolution models. The time convolution model is mainly used for modeling and analyzing time-dimension signal or sequence data, rather than traditional spatial image. Further, since the quality assessment in the present application is performed on a SAR image, and the essence of the spatial information of the SAR image is converted from time sequence signal, compared with the convolution neural network and Vit model suitable for processing spatial information in the prior art, the plurality of time convolution models are used to extract features of the SAR image in the present application, which can maximize the accuracy of feature extraction.

[0016] Further, the technical problem that the quality assessment of the SAR image is inaccurate if a model suitable for processing spatial analysis data is used to assess the quality of the SAR image due to the essence of the spatial information of the SAR image being converted from time sequence signal is solved. BRIEF DESCRIPTION OF DRAWINGS

[0017] The accompanying drawings, which are included to provide a further understanding of the present disclosure and constitute a part of this application, illustrate certain illustrative embodiments of the present disclosure and are used to explain the present disclosure, but do not limit the present disclosure. In the drawings: Figure 1is a schematic diagram of a communication connection system of a SAR satellite and a terminal device according to the embodiment 1 of the present application; Figure 2A is a schematic diagram of a hardware architecture of a SAR satellite according to the embodiment 1 of the present application; Figure 2B is a schematic diagram of a hardware architecture of a terminal device according to the embodiment 1 of the present application; Figure 3 is a flow chart of a quality evaluation method of a SAR image according to the embodiment 1 of the present application; Figure 4 is a schematic diagram of a quality evaluation model according to the embodiment 1 of the present application; Figure 5A is a schematic diagram of a convolution process of a time convolution model TCN-1 according to the embodiment 1 of the present application; Figure 5B is a schematic diagram of a convolution process of a time convolution model TCN-2 according to the embodiment 1 of the present application; Figure 6 is a schematic diagram of a quality evaluation device of a SAR image according to the embodiment 2 of the present application; and Figure 7 is a schematic diagram of a quality evaluation device of a SAR image according to the embodiment 3 of the present application. DETAILED DESCRIPTION

[0018] In order to make the technical personnel in the art better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, not all. Based on the embodiments in the present disclosure, all other embodiments obtained by those of ordinary skill in the art without creative labor should belong to the scope of protection of the present disclosure.

[0019] It should be noted that the terms "first", "second" and the like in the specification and claims of the present disclosure and the above-described drawings are used to distinguish similar objects, and do not necessarily indicate a specific order or a chronological sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present disclosure described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0020] Embodiment 1 According to this embodiment, a method embodiment for SAR image quality assessment is provided. It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system, such as a set of computer-executable instructions. Moreover, although a logical order is shown in the flowcharts, in some cases, the steps shown or described may be executed in a different order than shown.

[0021] Figure 1 A schematic diagram of a communication connection system between a SAR satellite and a terminal device according to this embodiment is shown. The system includes a terminal device 10, a ground system 20, and a SAR satellite 30. The terminal device 10 transmits a command to the SAR satellite 30 via the ground system 20 to perform a quality assessment on a SAR image corresponding to a target area. The processor of the SAR satellite 30 receives and responds to the command, acquires a SAR image corresponding to the target area, and inputs the SAR image into a pre-trained quality assessment model. The quality assessment model then outputs a quality assessment result corresponding to the SAR image.

[0022] Figure 2A It further shows Figure 1 Schematic diagram of the hardware architecture of the SAR satellite 30. Figure 2A As shown, the SAR satellite 30 includes an integrated electronic system, which includes: a processor, a memory, a bus management module and a communication interface. The memory is connected to the processor, so that the processor can access the memory, read the program instructions stored in the memory, read data from the memory or write data to the memory. The bus management module is connected to the processor and is also connected to a bus such as a CAN bus. The processor can communicate with the onboard peripherals connected to the bus through the bus managed by the bus management module. In addition, the processor is also connected to devices such as cameras, star sensors, measurement and control transponders, and data transmission equipment via the communication interface. It can be understood by those skilled in the art that Figure 2A The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 2A More or fewer components than shown, or with Figure 2A Different configurations shown.

[0023] Figure 2B It further shows Figure 1 Schematic diagram of the hardware architecture of the terminal device 10. Figure 2BAs shown, the terminal device 10 can include one or more processors (the processor can include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory for storing data, a transmission device for communication function, and an input / output interface. The memory, the transmission device, and the input / output interface are connected with the processor through a bus. In addition, a display, a keyboard, and a cursor control device connected with the input / output interface can be included. Those skilled in the art can understand that, Figure 2B The structure shown is only schematic, and does not limit the structure of the electronic device described above. For example, the ground system can include more or fewer components than those shown in Figure 2B or have a different configuration than Figure 2B that shown in the figures.

[0024] It should be noted that, Figure 2A and Figure 2B one or more processors and / or other data processing circuitry shown in the figures can be referred to herein generally as "data processing circuitry". The data processing circuitry can be embodied in whole or in part as software, hardware, firmware, or any combination thereof. Furthermore, the data processing circuitry can be a single standalone processing module, or it can be incorporated in whole or in part within any other component of the computing device. As referred to in the embodiments of the present disclosure, the data processing circuitry serves as a processor to control, for example, the selection of the variable resistance terminal path connected with the interface.

[0025] Figure 2A and Figure 2B The memory shown in the figures can be used to store software programs and modules of application software, such as the program instructions / data storage means corresponding to the SAR image quality evaluation method in the embodiments of the present disclosure. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, implements the SAR image quality evaluation method of the application program described above. The memory can include a high-speed random access memory, and can also include a non-volatile memory, such as one or more magnetic storage devices, flash memories, or other non-volatile solid-state memories.

[0026] It should be noted that in some optional embodiments, the above-mentioned Figure 2A and Figure 2B The device shown in the figures can include hardware elements (including circuitry), software elements (including computer code stored on a computer readable medium), or a combination of both hardware and software elements. It should be noted that, Figure 2A and Figure 2B is only one example of a particular concrete instance, and is intended to show the types of components that can be present in the above-described device.

[0027] In the above operating environment, according to a first aspect of the embodiment, a quality evaluation method of a SAR image is provided, the method is implemented by the SAR satellite 30 shown in Figure 2A Figure 3 A flowchart of the method is shown, referring to Figure 3 The method comprises: S302: collecting a SAR image corresponding to a target region by using a SAR satellite, wherein the abscissa of the SAR image represents a time sequence related azimuth, the ordinate of the SAR image represents slant range, and the pixel value of the SAR image represents the backscattering intensity corresponding to the azimuth and the slant range; S304: inputting the SAR image into a pre-trained quality evaluation model, and determining a plurality of first feature maps corresponding to the SAR image by using each time convolution model in the quality evaluation model, wherein the time convolution model comprises a plurality of layers of time convolution networks, and the receptive field of each layer of time convolution network is different; and S306: performing quality evaluation on the SAR image corresponding to the target region based on the plurality of first feature maps and by using the quality evaluation model.

[0028] Specifically, first, the operator uses the terminal device 10 and sends an instruction to the SAR satellite 30 through the ground system 20 to evaluate the quality of the SAR image corresponding to the target region. The processor of the SAR satellite 30 receives and responds to the instruction to collect the SAR image corresponding to the target region (S302). Wherein, the flight direction of the SAR satellite 30 is the azimuth, and each row in the SAR image represents the signal intensity reflected back by the SAR satellite 30 at the same slant range and different azimuths. And all the pixel blocks in this row correspond to the same slant range between the SAR satellite and the target region. The side view direction of the SAR satellite 30 is the slant range, and each column in the SAR image represents the signal intensity reflected back at the same azimuth and different slant ranges. And all the pixel blocks in this column correspond to the same azimuth between the SAR satellite and the target region.

[0029] In addition, in the case of determining the abscissa corresponding to the SAR image and the ordinate corresponding to the SAR image, the signal intensity of the electromagnetic wave backscattering corresponding to the unique azimuth and the slant range is determined. In actual cases, the SAR image is a grayscale image, and the pixel value in the SAR image is equivalent to the signal intensity of the electromagnetic wave backscattering corresponding to the unique azimuth and the slant range. That is, the greater the backscattering intensity, the greater the pixel value; the smaller the backscattering intensity, the smaller the pixel value.

[0030] For example, the azimuth and the slant range in the SAR image are determined. And each combination of azimuth and slant range determines a unique pixel value. Wherein,​ corresponding to the pixel value corresponding to the pixel value corresponding to the pixel value corresponding to the pixel value corresponding to the pixel value corresponding to the pixel value corresponding to the pixel value corresponding to the pixel value corresponding to the pixel value corresponding to the pixel value corresponding to the pixel value corresponding to the pixel value corresponding to the pixel value corresponding to the pixel value corresponding to the pixel value corresponding to the pixel value corresponding to the pixel value corresponding to the pixel value corresponding to the pixel value corresponding to the pixel value corresponding to the pixel value corresponding to the pixel value corresponding to the pixel value corresponding to the pixel value

[0031] corresponding to the azimuth direction corresponding to the time corresponding to the time corresponding to the time vector corresponding to the time vector corresponding to the time vector corresponding to the time corresponding to the time corresponding to the time vector corresponding to the time vector corresponding to the time vector corresponding to the time corresponding to the time corresponding to the time vector corresponding to the time vector corresponding to the time vector

[0032] Thus, the SAR image can be actually represented by a plurality of time vector sequences. Each of the time vectors is used to indicate a distribution of pixel values corresponding to different slant ranges at a corresponding time. Each element in the time vector is used to indicate a pixel value corresponding to a different slant range.

[0033] That is, the essence of the spatial information of the SAR image is converted from the time sequence signal, and the SAR image can be represented by a plurality of time vector sequences.

[0034] Thus, the SAR satellite 30 inputs the SAR image into the pre-trained quality assessment model, and determines a plurality of first feature maps corresponding to the SAR image by using each temporal convolution model in the quality assessment model (S304).Figure 4 is a schematic diagram of a quality assessment model according to an embodiment of the present application. Referring to Figure 4 , the quality assessment model includes a plurality of temporal convolution models TCN-1-TCN-U, and each temporal convolution model includes a plurality of temporal convolution networks. For example, the temporal convolution model TCN-1 includes a plurality of temporal convolution networks, and the convolution kernel size of each layer of the temporal convolution networks is the same, but the step size of sliding on the SAR image is different, and the receptive field formed by the dilated convolution is also different. That is, the receptive field is expanded by using the interval sampling mode between each layer of the temporal convolution networks to capture long-term dependencies (i.e., dilated convolution).

[0035] Figure 5A is a schematic diagram of a convolution process of the temporal convolution model TCN-1 according to an embodiment of the present application. Referring to Figure 5A , the convolution kernel size of the first layer of the temporal convolution network (input layer) of the temporal convolution model TCN-1 is 3. The convolution kernel size of the second layer of the temporal convolution network (hidden layer) of the temporal convolution model TCN-1 is 3, and the step size is 1. The convolution kernel size of the third layer of the temporal convolution network (hidden layer) of the temporal convolution model TCN-1 is 3, and the step size is 2. That is, the receptive field of the third layer of the temporal convolution network (hidden layer) is 7, the receptive field of the second layer of the temporal convolution network (hidden layer) is 5, and the receptive field of the first layer of the temporal convolution network (input layer) is 3. The receptive field of the third layer of the temporal convolution network is increased compared to the receptive field of the second layer of the temporal convolution network, and the receptive field of the second layer of the temporal convolution network is increased compared to the receptive field of the first layer of the temporal convolution network.

[0036] Figure 5B is a schematic diagram of a convolution process of the temporal convolution model TCN-2 according to an embodiment of the present application. Referring to Figure 5B , and wherein the convolution kernel size of the first layer of the temporal convolution network (input layer) of the temporal convolution model TCN-2 is 3, and the step size is 1. The convolution kernel size of the second layer of the temporal convolution network (hidden layer) of the temporal convolution model TCN-2 is 3, and the step size is 2. The convolution kernel size of the second layer of the temporal convolution network (hidden layer) of the temporal convolution model TCN-3 is 3, and the step size is 3. That is, the receptive field of the third layer of the temporal convolution network (hidden layer) is 9, the receptive field of the second layer of the temporal convolution network (hidden layer) is 7, and the receptive field of the first layer of the temporal convolution network (input layer) is 5. The receptive field of the third layer of the temporal convolution network is increased compared to the receptive field of the second layer of the temporal convolution network, and the receptive field of the second layer of the temporal convolution network is increased compared to the receptive field of the first layer of the temporal convolution network.

[0037] Comparison Figure 5A and Figure 5B It can be seen that the receptive fields of each layer of the temporal convolution networks of the first temporal convolution model and the second temporal convolution model are not the same.

[0038] Therefore, unlike the prior art which uses convolutional neural networks or ViT to extract features from SAR images, since the present application uses a time convolution model suitable for processing time series signals to process SAR images, it is more accurate than traditional methods.

[0039] Finally, in the case where each time convolution model outputs a plurality of first feature maps corresponding to the target region, the quality of the SAR image corresponding to the target region is evaluated based on the plurality of first feature maps and using a quality evaluation model (S306). Therefore, the operator can determine the quality analysis result corresponding to the SAR image through the terminal device 10, and evaluate the quality of the SAR image based on the quality analysis result.

[0040] As described in the background, in the quality evaluation of SAR images, convolutional neural networks (CNN) or Vit are usually used to process SAR images. However, convolutional neural networks (CNN) and Vit perform well in processing spatial distribution data (e.g., optical remote sensing images), but have certain limitations in processing SAR images which are essentially time series signals. For example, since the pixel values of optical remote sensing images directly reflect the spatial radiation characteristics of ground objects, using convolutional neural networks (CNN) can effectively extract features of adjacent pixels belonging to the same ground object, or using the self-attention mechanism of ViT can effectively extract ground object distribution features across the entire image.

[0041] However, since the spatial information of SAR images is essentially converted from time series signals, directly using convolutional neural networks (CNN) or Vit to process SAR images for evaluating the quality of SAR images may not be accurate.

[0042] Therefore, the present application discloses a quality evaluation model, and the quality evaluation model includes a plurality of time convolution models. The time convolution model is mainly used for modeling and analyzing time dimension signal or sequence data, rather than traditional spatial images. Further, since the present application is to evaluate the quality of SAR images, and the spatial information of SAR images is essentially converted from time series signals, compared with the prior art which uses convolutional neural networks and Vit models suitable for processing spatial information, the present application uses a plurality of time convolution models to extract features from SAR images, which can maximize the accuracy of feature extraction.

[0043] Further, it solves the technical problem that the evaluation is not accurate in the prior art due to the fact that the spatial information of SAR images is essentially converted from time series signals, and if a model suitable for processing spatial analysis data is used to evaluate the quality of SAR images, the evaluation may not be accurate.

[0044] Optionally, the operation of determining the plurality of first feature maps corresponding to the SAR image by using each time convolution model in the quality evaluation model comprises: inputting the SAR image into each time convolution model respectively; and performing dilated convolution on the SAR image in the column direction of the SAR image by using the convolution kernel corresponding to each layer of time convolution network in the time convolution model, and generating the first feature map corresponding to each time convolution model.

[0045] Specifically, referring to FIG. 1, taking the first time convolution model TCN-1 as an example, the SAR image is input into the first time convolution model TCN-1, and the convolution kernel of each layer of time convolution network in the first time convolution model TCN-1 is used to perform dilated convolution on the SAR image in the column direction of the SAR image with different receptive fields. Thus, the first feature map corresponding to the time convolution model TCN-1 is generated. Figure 5A

[0046] For example, the convolution kernel size of the first layer of time convolution network (input layer) of the first time convolution model TCN-1 is 3, and the convolution kernel in the first layer of time convolution network is used to perform convolution on each pixel block on the SAR image in the column direction of the SAR image.

[0047] The convolution kernel size of the second layer of time convolution network (hidden layer) of the first time convolution model TCN-1 is 3, the step is 1, and the receptive field is 5. The convolution kernel in the second layer of time convolution network is used to perform convolution on each pixel block on the SAR image in the column direction of the SAR image.

[0048] The convolution kernel size of the third layer of time convolution network (hidden layer) of the first time convolution model TCN-1 is 3, the step is 1, and the receptive field is 7. The convolution kernel in the second layer of time convolution network is used to perform convolution on each pixel block on the SAR image in the column direction of the SAR image.

[0049] Thus, the receptive field of the third layer of time convolution network is increased compared with the receptive field of the second layer of time convolution network, and the receptive field of the second layer of time convolution network is increased compared with the receptive field of the first layer of time convolution network.

[0050] By analogy, the receptive field of the last layer of time convolution network of the first time convolution model TCN-1 is increased compared with the receptive field of the time convolution network of the layer before the last layer.

[0051] ​Therefore, when the receptive fields of each layer of the temporal convolutional network are different and the receptive fields are constantly increasing, the shallow temporal convolutional network with a small receptive field can capture local timing patterns, and the deep temporal convolutional network with a large receptive field can identify global periodic change trends. Therefore, compared with traditional convolutional neural networks, it can extract features more accurately and finally output the first feature map.

[0052] Similarly, other temporal convolution models TCN-2~TCN-U can output the corresponding first feature map.

[0053] Therefore, multiple time convolution models TCN-1 to TCN-U can output multiple first feature maps corresponding to the SAR image.

[0054] Optionally, the quality assessment model includes multiple transformer models corresponding to each temporal convolution model, a convolutional neural network, an MLP model, a fully connected layer and a softmax classifier corresponding to each transformer model, and performs a quality assessment on the SAR image corresponding to the target area based on multiple first feature maps, including: inputting the feature map corresponding to each temporal convolution model into the corresponding transformer model, and outputting the second feature map corresponding to each first feature map; inputting each second feature map into the convolutional neural network, and outputting the third feature map corresponding to each second feature map; inputting multiple third feature maps into the MLP model, determining the feature vector corresponding to each third feature map, splicing the feature vectors, and generating a fused feature vector; and inputting the fused feature vector into the fully connected layer and the softmax classifier, and performing quality assessment on the SAR image based on the output results.

[0055] Specifically, refer to Figure 4 As shown in the figure, first, when each temporal convolutional model TCN-1 to TCN-U outputs a first feature map corresponding to the SAR image, each first feature map is input into the corresponding transformer model, and the transformer model then outputs a second feature map corresponding to each first feature map. When the first feature map is input into the transformer model, the first feature map is flattened into a corresponding sequence of time series vectors. Furthermore, when the transformer model uses a self-attention mechanism to process the feature vectors and generate a sequence of feature vectors, the sequence of feature vectors is reshaped into a second feature map.

[0056] Then, the quality evaluation model respectively inputs each second feature map into a convolutional neural network, and outputs a third feature map corresponding to each second feature map. Further, the quality evaluation model inputs multiple third feature maps into an MLP model, and determines a feature vector corresponding to each third feature map. Then, each feature vector is spliced to generate a fusion feature vector. Finally, the quality evaluation model inputs the fusion feature vector into a fully connected layer and a softmax classifier, thereby outputting a probability corresponding to each quality category.

[0057] Optionally, based on the output result, the operation of quality evaluating the SAR image includes: determining a quality evaluation result corresponding to the SAR image based on the probability of the SAR image in each quality category output by the softmax classifier, wherein the quality category includes high quality, general quality and low quality.

[0058] Specifically, the quality category includes high quality, general quality and low quality. And in the case that the softmax classifier outputs the probability of the SAR image in each quality category, the maximum probability value in the probability corresponding to each quality category is determined. And the quality category corresponding to the maximum probability value is determined as the quality evaluation result corresponding to the SAR image. For example, the probability value corresponding to high quality is 70%, the probability value corresponding to general quality is 20%, and the probability value corresponding to low quality is 10%.

[0059] Thus, through the above operation, the quality of the SAR image can be accurately analyzed.

[0060] In addition, the SAR satellite 30 needs to train the quality evaluation model in advance before determining the quality of the SAR image corresponding to the target area by using the quality evaluation model. The specific training steps include: Firstly, the SAR satellite 30 acquires SAR image samples corresponding to each sample area. Each SAR image sample has a label for indicating the corresponding quality category, and the label includes high quality, general quality and low quality.

[0061] Then, the SAR satellite constructs a quality evaluation model. The quality evaluation model includes multiple temporal convolution models, multiple transformer models, multiple convolutional neural networks, an MLP model, a fully connected layer and a softmax classifier.

[0062] After that, multiple labeled SAR image samples are input into the quality evaluation model, and the quality evaluation model is trained.

[0063] Thus, according to the first aspect of the present embodiment, the technical effect of maximizing the accuracy of feature extraction from the SAR image, and thus maximizing the accuracy of quality assessment of the SAR image, is achieved.

[0064] Further, referring to Figure 1 As shown in Fig. 2, according to the second aspect of the present embodiment, a storage medium is provided. The storage medium includes a stored program, wherein the program is executed by a processor to perform any of the above methods when the program is run.

[0065] Thus, according to the present embodiment, the technical effect of maximizing the accuracy of feature extraction from the SAR image, and thus maximizing the accuracy of quality assessment of the SAR image, is achieved.

[0066] It should be noted that, for the above-mentioned method embodiments, in order to simply describe, they are all expressed as a series of action combinations, but those skilled in the art should know that the present application is not limited by the action sequence described, because according to the present application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should know that the embodiments described in the specification all belong to preferred embodiments, and the actions and modules involved are not necessarily essential to the present application.

[0067] From the above description of the embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be realized by means of software and the necessary general hardware platform, of course, it can also be realized by hardware, but in many cases the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions to make a terminal device (which can be a mobile phone, computer, server, or network device, etc.) execute the methods described in the embodiments of the present application.

[0068] Embodiment 2 Figure 6The device 600 for quality evaluation of a SAR image according to the embodiment is shown, which corresponds to the method according to the embodiment 1. As shown in reference 6, the device 600 comprises: a SAR image acquisition module 610, configured to acquire a SAR image corresponding to a target area by using a SAR satellite, wherein the abscissa of the SAR image represents a time sequence related azimuth, the ordinate of the SAR image represents a slant range, and the pixel value of the SAR image represents a backscattering intensity corresponding to the azimuth and the slant range; a first feature map determination module 620, configured to input the SAR image into a pre-trained quality evaluation model, and determine a plurality of first feature maps corresponding to the SAR image by using each time convolution model in the quality evaluation model, wherein the time convolution model comprises a plurality of time convolution networks, and the receptive field of each layer of the time convolution network is different; and a quality evaluation module 630, configured to perform quality evaluation on the SAR image corresponding to the target area based on the plurality of first feature maps and by using the quality evaluation model.

[0069] Optionally, the first feature map determination module 620 comprises: a SAR image input module, configured to input the SAR image into each time convolution model respectively; and a first feature map output module, configured to perform a dilated convolution on the SAR image in a column direction of the SAR image by using a convolution kernel corresponding to each layer of the time convolution network in the time convolution model, and generate a first feature map corresponding to each time convolution model.

[0070] Optionally, the quality evaluation model comprises a plurality of transformer models corresponding to each time convolution model respectively, a convolutional neural network corresponding to each transformer model respectively, an MLP model, a full connection layer and a softmax classifier, and the quality evaluation module 630 comprises: a second feature map output module, configured to input the feature map corresponding to each time convolution model into the corresponding transformer model respectively, and output a second feature map corresponding to each first feature map; a third feature map output module, configured to input each second feature map into the convolutional neural network respectively, and output a third feature map corresponding to each second feature map; a fused feature vector generation module, configured to input a plurality of third feature maps into the MLP model, determine a feature vector corresponding to each third feature map, splice each feature vector, and generate a fused feature vector; and a quality evaluation sub-module, configured to input the fused feature vector into the full connection layer and the softmax classifier, and perform quality evaluation on the SAR image based on an output result.

[0071] Optionally, the quality assessment submodule comprises a quality assessment result output module configured to determine a quality assessment result corresponding to the SAR image based on probabilities of the SAR image in each quality category output by the softmax classifier, wherein the quality categories comprise high quality, general quality, and low quality.

[0072] According to the present embodiment, the technical effect of maximizing the accuracy of feature extraction of the SAR image and ensuring accurate quality assessment of the SAR image is achieved.

[0073] Embodiment 3 Figure 7 A SAR image quality assessment device 700 according to the present embodiment is shown, which corresponds to the method according to Embodiment 1. Referring to Figure 7 As shown, the device 700 comprises a processor 710 and a memory 720 connected to the processor 710, configured to provide the processor 710 with instructions to process the following processing steps: collecting a SAR image corresponding to a target region by a SAR satellite, wherein the abscissa of the SAR image represents a time sequence related azimuth, the ordinate of the SAR image represents a slant range, and the pixel value of the SAR image represents a backscattering intensity corresponding to the azimuth and the slant range; inputting the SAR image into a pre-trained quality assessment model, and determining a plurality of first feature maps corresponding to the SAR image by using each time convolution model in the quality assessment model, wherein the time convolution model comprises a plurality of layers of time convolution networks, and the receptive field of each layer of time convolution network is different; and performing quality assessment on the SAR image corresponding to the target region based on the plurality of first feature maps and by using the quality assessment model.

[0074] Optionally, the operation of determining a plurality of first feature maps corresponding to the SAR image by using each time convolution model in the quality assessment model comprises: inputting the SAR image into each time convolution model respectively; and performing dilated convolution on the SAR image in the column direction by using the convolution kernel corresponding to each layer of time convolution network in the time convolution model with different receptive fields, and generating a first feature map corresponding to each time convolution model.

[0075] Optionally, the quality assessment model comprises a plurality of transformer models corresponding to the respective time convolution models, a convolutional neural network corresponding to the respective transformer models, an MLP model, a full connection layer and a softmax classifier, and based on the plurality of first feature maps, the operation of performing quality assessment on the SAR image corresponding to the target region comprises: inputting the feature maps corresponding to the respective time convolution models into the corresponding transformer models respectively, and outputting second feature maps corresponding to the respective first feature maps; inputting the respective second feature maps into the convolutional neural network respectively, and outputting third feature maps corresponding to the respective second feature maps; inputting the plurality of third feature maps into the MLP model, determining feature vectors corresponding to the respective third feature maps, splicing the respective feature vectors, and generating a fusion feature vector; and inputting the fusion feature vector into the full connection layer and the softmax classifier, and based on the output result, performing quality assessment on the SAR image.

[0076] Optionally, based on the output result, the operation of performing quality assessment on the SAR image comprises: determining a quality assessment result corresponding to the SAR image based on the probabilities of the SAR image in the respective quality categories output by the softmax classifier, wherein the quality categories comprise high quality, general quality and low quality.

[0077] Therefore, according to the embodiment, the technical effect of being able to maximize the accuracy of feature extraction on the SAR image, and further being able to accurately perform quality assessment on the SAR image is achieved.

[0078] The above-mentioned serial numbers of the embodiments of the application are only for description, and do not represent the advantages or disadvantages of the embodiments.

[0079] In the above-mentioned embodiments of the application, the description of each embodiment has its own emphasis, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0080] In several embodiments provided in the present application, it should be understood that the disclosed technical content can be implemented by other ways. Among them, the above-mentioned device embodiments are only schematic, for example, the division of the units is only a logical function division, and actual implementation can have another division way, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point, the coupling or direct coupling or communication connection between the displayed or discussed each other can be through some interface, indirect coupling or communication connection between units or modules, which can be electrical or other forms.

[0081] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed on multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0082] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0083] The integrated unit, if realized in the form of a software functional unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical scheme of the present application or the part of the present application which contributes to the prior art or the whole or part of the technical scheme can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The foregoing storage medium includes: a U disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a mobile hard disk, a magnetic disk or an optical disk, and various program code storage media.

[0084] The above is only the preferred embodiment of the present application. It should be noted that for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, and these improvements and refinements should be considered as the protection scope of the present application.

Claims

1. A SAR image quality assessment method, characterized in that: include: Collecting a SAR image corresponding to the target area using a SAR satellite, wherein the abscissa of the SAR image represents an azimuth related to a time series, the ordinate of the SAR image represents a slant distance, and a pixel value of the SAR image represents a backscatter intensity corresponding to the azimuth and the slant distance; Inputting the SAR image into a pre-trained quality assessment model, and determining a plurality of first feature maps corresponding to the SAR image using respective temporal convolutional models in the quality assessment model, wherein the temporal convolutional model comprises a multi-layer temporal convolutional network, and each layer of the temporal convolutional network has a different receptive field; as well as Based on the multiple first feature maps and using the quality assessment model, quality assessment is performed on the SAR image corresponding to the target area.

2. The method according to claim 1, characterized in that The operation of determining a plurality of first feature maps corresponding to the SAR image using each temporal convolution model in the quality assessment model includes: Inputting the SAR images into respective temporal convolution models; and Using the convolution kernels corresponding to the temporal convolution networks of each layer in the temporal convolution model, the SAR image is subjected to dilated convolution with different receptive fields in the column direction of the SAR image, and a first feature map corresponding to each temporal convolution model is generated.

3. The method according to claim 1, characterized in that The quality assessment model includes a plurality of transformer models corresponding to the respective temporal convolution models, a convolutional neural network, an MLP model, a fully connected layer, and a softmax classifier corresponding to the respective transformer models, and an operation of performing a quality assessment on the SAR image corresponding to the target area based on the plurality of first feature maps, including: Inputting the first feature maps corresponding to the respective temporal convolutional models into the corresponding transformer models, and outputting the second feature maps corresponding to the respective first feature maps; Inputting each second feature map into the convolutional neural network respectively, and outputting a third feature map corresponding to each second feature map; Inputting multiple third feature maps into the MLP model, determining feature vectors corresponding to each third feature map, concatenating the feature vectors, and generating a fused feature vector; and The fused feature vector is input into the fully connected layer and the softmax classifier, and the quality of the SAR image is evaluated based on the output results.

4. The method according to claim 3, characterized in that The operation of performing quality assessment on the SAR image based on the output result includes: A quality assessment result corresponding to the SAR image is determined based on the probability of the SAR image in each quality category output by the softmax classifier, where the quality categories include high quality, average quality, and low quality.

5. A storage medium, characterized in that The storage medium includes a stored program, wherein when the program is run, the processor executes the method according to any one of claims 1 to 4.

6. A SAR image quality assessment device, characterized in that: include: A SAR image acquisition module, configured to acquire a SAR image corresponding to the target area using a SAR satellite, wherein the abscissa of the SAR image represents an azimuth direction related to a time series, the ordinate of the SAR image represents a slant distance, and the pixel value of the SAR image represents a backscattering intensity corresponding to the azimuth direction and the slant distance; a first feature map determination module, configured to input the SAR image into a pre-trained quality assessment model and determine a plurality of first feature maps corresponding to the SAR image using respective temporal convolutional models in the quality assessment model, wherein the temporal convolutional model comprises a multi-layer temporal convolutional network, and each layer of the temporal convolutional network has a different receptive field; as well as A quality assessment module is configured to perform quality assessment on the SAR image corresponding to the target area based on the multiple first feature maps and using a quality assessment model.

7. The device according to claim 6, characterized in that The first feature map determining module includes: A SAR image input module, configured to input the SAR image into each temporal convolution model respectively; and The first feature map output module is used to use the convolution kernels corresponding to the temporal convolution networks of each layer in the temporal convolution model to perform dilated convolution on the SAR image with different receptive fields in the column direction of the SAR image, and generate a first feature map corresponding to each temporal convolution model.

8. The device according to claim 6, characterized in that The quality assessment model includes multiple transformer models corresponding to the respective temporal convolution models, a convolutional neural network, an MLP model, a fully connected layer, and a softmax classifier corresponding to each transformer model, and the quality assessment module includes: A second feature map output module, configured to input the feature maps corresponding to the respective temporal convolutional models into the corresponding transformer models, and output a second feature map corresponding to each first feature map; a third feature map output module, configured to input each second feature map into the convolutional neural network, and output a third feature map corresponding to each second feature map; a fused feature vector generation module, configured to input a plurality of third feature maps into the MLP model, determine a feature vector corresponding to each third feature map, concatenate the feature vectors, and generate a fused feature vector; and The quality assessment submodule is used to input the fused feature vector into the fully connected layer and the softmax classifier, and perform quality assessment on the SAR image based on the output results.

9. The device according to claim 8, characterized in that The quality assessment submodule includes: A quality assessment result output module is configured to determine a quality assessment result corresponding to the SAR image based on the probability of the SAR image in each quality category output by the softmax classifier, wherein the quality categories include high quality, average quality, and low quality.

10. A SAR image quality assessment device, characterized in that: include: processor; as well as A memory, connected to the processor, configured to provide the processor with instructions for processing the following processing steps: Collecting a SAR image corresponding to the target area using a SAR satellite, wherein the abscissa of the SAR image represents an azimuth related to a time series, the ordinate of the SAR image represents a slant distance, and a pixel value of the SAR image represents a backscatter intensity corresponding to the azimuth and the slant distance; Inputting the SAR image into a pre-trained quality assessment model, and determining a plurality of first feature maps corresponding to the SAR image using respective temporal convolutional models in the quality assessment model, wherein the temporal convolutional model comprises a multi-layer temporal convolutional network, and each layer of the temporal convolutional network has a different receptive field; as well as Based on the multiple first feature maps and using the quality assessment model, quality assessment is performed on the SAR image corresponding to the target area.

Images