SAR image quality evaluation method based on target detection and recognition

Abstract

The invention relates to an SAR image quality evaluation method based on target detection and recognition, and the method comprises the steps: obtaining an SAR image set containing a plurality of SAR images, and carrying out the target detection and recognition of the SAR images, so as to obtain the target detection and recognition accuracy; calculating an image ambiguity evaluation index of the SAR image; calculating an image noise level evaluation index of the SAR image; and establishing a relevance model of the image ambiguity evaluation index, the image noise level evaluation index and the target detection and recognition accuracy based on a linear regression method to obtain SAR image quality parameters for evaluating the SAR image quality. According to the SAR image quality evaluation method based on target detection and recognition provided by the invention, the image quality is evaluated without a reference image, and the limitation that a high-quality reference image needs to be acquired is avoided; target detection and recognition and SAR image quality evaluation are associated, SAR image quality parameters are formed based on multiple evaluation indexes, and the evaluation capability and accuracy of the SAR image are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of SAR image quality evaluation, and in particular to a SAR image quality evaluation method based on target detection and recognition. Background Art

[0002] In recent years, Synthetic Aperture Radar (SAR) has gradually matured. SAR is an active microwave remote sensing technology that uses radar signals to perform high-resolution imaging of the ground by being carried on mobile platforms such as satellites, aircraft, or drones. Due to the greatly improved imaging quality, it is widely used in various fields. Compared with optical remote sensing, SAR has strong penetration and can avoid the influence of factors such as climate, light, and trees. It has the characteristics of all-weather and all-day operation. This advantage has played a great auxiliary role in the military field, such as monitoring military targets such as tanks, aircraft, and ships. It also brings convenience to the civilian field, such as observing crop growth and resource exploration.

[0003] SAR image quality is affected by many factors, such as blur and noise levels. For example, images with low blur contain more useful information and detail, while images with high blur have the opposite effect. Accurately evaluating SAR image quality facilitates adjustments to platform parameters during imaging, which in turn influence image parameters and achieve better detection and recognition performance.

[0004] In related technologies, methods for evaluating image quality have limitations. If image quality is evaluated using a reference image, high-quality images and degraded images need to be compared, and high-quality reference images are often difficult to obtain. If image quality is evaluated without a reference image, such as from a single aspect such as image contrast, brightness, clarity, and noise, the accuracy is low. Summary of the Invention

[0005] The present application provides a SAR image quality assessment method based on target detection and recognition, which solves the technical problem that image quality assessment methods in related technologies have limitations.

[0006] The present invention provides a SAR image quality assessment method based on target detection and recognition, which includes the following steps: Acquire a SAR image set comprising a plurality of SAR images, and perform target detection and recognition on the SAR images to obtain a target detection and recognition accuracy rate; Calculating an image fuzziness evaluation index of the SAR image; Calculating an image noise level evaluation index of the SAR image; Based on the linear regression method, a correlation model between the image blur evaluation index, the image noise level evaluation index and the target detection and recognition accuracy is established to obtain a SAR image quality parameter for evaluating the SAR image quality.

[0007] In one embodiment, establishing a correlation model between the image blur evaluation index, the image noise level evaluation index, and the target detection and recognition accuracy based on a linear regression method to obtain a SAR image quality parameter for evaluating the SAR image quality includes: Using the image blur evaluation index and the image noise level evaluation index as predictor variables and the target detection and recognition accuracy rate as the response variable, linear regression fitting is performed to obtain a linear regression model; The linear regression model is expressed as: ; ; in, BAH is the image blur evaluation index, LS and WN is the image noise level evaluation index, P The target detection and recognition accuracy; The SAR image quality parameters are formed by the linear regression model OQ , expressed as: .

[0008] In one embodiment, the image blur evaluation index is a blur evaluation index based on the image histogram. BAH .

[0009] In one embodiment, the fuzziness evaluation index based on the image histogram is calculated BAH include: Statistically calculate the grayscale histogram of the SAR image and the distribution probability of each grayscale level ; The calculation formula is: ; in, The gray value in the SAR image is equal to x The number of pixels, is the total number of pixels in the SAR image; Calculate the weight of each grayscale value ; The calculation formula is: ; in, is the grayscale mean of the image; Calculate the degree evaluation index BAH ; The calculation formula is: .

[0010] In one embodiment, the image noise level evaluation index includes a noise level evaluation index based on the local standard deviation. LS , Noise level evaluation index based on wavelet transform WN .

[0011] In one embodiment, the noise level evaluation index based on the local standard deviation is calculated LS include: Divide a SAR image into several pixel-sized image blocks and calculate the local mean of each image block. E and local standard deviation S； The calculation formula is: , ; in, Indicates the i The total number of pixels in the image block, Indicates the i In the image block j Image Gray value of pixel; Divide the area between the maximum local standard deviation and the minimum local standard deviation by a set multiple into a number of equally spaced intervals, the local standard deviations of all the image blocks fall into the corresponding intervals, and count the number of image blocks in each interval; Calculate the signal-to-noise ratio to get LS ; The calculation formula is: ; in, N is the total number of image blocks, M is the number of image blocks in the interval with the largest number of image blocks, arrive q Number the image blocks that fall into the above interval.

[0012] In one embodiment, the noise level evaluation index based on wavelet transform is calculated WN include: The calculation formula is: ; in, Refers to the set frequency information in the diagonal direction of the image.

[0013] In one embodiment, acquiring a SAR image set comprising a plurality of SAR images includes: Get SAR original image; increasing the image blur and noise level of the SAR original image to obtain a plurality of degraded images; The SAR original image and the degraded image are used as the SAR images to form the SAR image set.

[0014] In one embodiment, performing target detection and recognition on the SAR image to obtain a target detection and recognition accuracy rate includes: Dividing the SAR images of different image qualities into a training set and a test set; The target detection and recognition model is trained and tested, and the target detection and recognition performance is described by the target detection and recognition accuracy.

[0015] In one embodiment, the target detection and recognition model is a YOLO-V8 model based on a residual network.

[0016] The beneficial effects of the technical solutions provided in the embodiments of the present application include: The present application provides a SAR image quality evaluation method based on target detection and recognition. A correlation model between an image blur evaluation index, an image noise level evaluation index, and target detection and recognition accuracy is established based on a linear regression method to obtain SAR image quality parameters for evaluating SAR image quality. On the one hand, no reference image is required to evaluate image quality, thus avoiding the limitation of having to obtain high-quality reference images. On the other hand, the method links computer vision tasks (i.e., target detection and recognition) with SAR image quality evaluation, and forms reference SAR image quality parameters based on multiple evaluation indicators. This improves the evaluation capability and accuracy of SAR images, effectively predicts target recognition in SAR images, and provides screening guidance for high-quality SAR image datasets. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0018] Figure 1 The figure is a flowchart of the steps of a SAR image quality assessment method based on target detection and recognition in one embodiment of the present invention.

[0019] Figure 2 This is a partial target slice image of the FUSAR-ship image set in one embodiment of the present invention.

[0020] Figure 3 Schematic diagram of the change in loss during target detection model training in one embodiment of the present invention.

[0021] Figure 4 This is a deviation diagram of the target detection recognition accuracy predicted by the linear regression model and the recognition result of the target detection model in one embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to enable those skilled in the art to better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 this application.

[0023] The embodiments of the present application provide a SAR image quality assessment method based on target detection and recognition, which can solve the technical problem of limitations of image quality assessment methods in related technologies.

[0024] like Figure 1 As shown, Figure 1 The figure is a flowchart of the steps of a SAR image quality assessment method based on target detection and recognition in one embodiment of the present invention.

[0025] This embodiment provides a SAR image quality assessment method based on target detection and recognition, which includes the following steps: Step S1: obtaining a SAR image set including a plurality of SAR images, performing target detection and recognition on the SAR images to obtain a target detection and recognition accuracy rate; Step S2, calculating the image ambiguity evaluation index of the SAR image; Step S3, calculating the image noise level evaluation index of the SAR image; Step S4: establishing a correlation model between the image blur evaluation index, the image noise level evaluation index and the target detection and recognition accuracy based on a linear regression method to obtain a SAR image quality parameter for evaluating the SAR image quality.

[0026] This embodiment provides a SAR image quality evaluation method based on target detection and recognition. Based on the linear regression method, a correlation model between an image blur evaluation index, an image noise level evaluation index and the target detection and recognition accuracy is established to obtain SAR image quality parameters for evaluating SAR image quality. On the one hand, no reference image is required to evaluate image quality, thus avoiding the limitation of having to obtain high-quality reference images. On the other hand, the computer vision task (i.e., target detection and recognition) and SAR image quality evaluation are linked. Together, the SAR image quality parameters that can be used as reference are formed based on multiple evaluation indicators, which can improve the evaluation ability and accuracy of SAR images. It can also effectively predict the target recognition situation in SAR images and provide screening guidance for high-quality SAR image datasets.

[0027] Each step is described and explained in detail below.

[0028] In one embodiment, in step S1, step S11, obtaining a SAR image set including a plurality of SAR images, includes: Step S111: Acquire the original SAR image.

[0029] like Figure 2 As shown, Figure 2 This is a partial target slice image of the FUSAR-ship image set in one embodiment of the present invention.

[0030] Specifically, taking the FUSAR-Ship high-resolution sea surface ship image set as an example, 100 typical FUSAR-ship images are selected as SAR original images.

[0031] Step S112: increasing the image blur and noise level of the original SAR image to obtain a plurality of degraded images.

[0032] Specifically, a FUSAR-ship image (SAR original image) was selected and its resolution was reduced to 1.2 times, 1.4 times, 1.6 times, 1.8 times, and 8.0 times of the SAR original image; Gaussian noise was superimposed on the SAR original image with the values ​​of 5, 10, 15, 20, and 25 dBW, respectively. When the noise was combined in pairs, 35 degraded images could be generated for each SAR original image.

[0033] Step S113: Using the original SAR image and the degraded image as SAR images, a SAR image set is formed.

[0034] 100 original SAR images correspond to 3500 degraded images, forming a SAR image set containing 3600 SAR images.

[0035] Through the above scheme, by combining resolution reduction (such as 1.2~8.0 times) and superimposing Gaussian noise (such as 5~25dBW), the blur and noise commonly seen in real scenes are simulated to artificially generate large-scale images of different qualities. SAR images, significantly reducing data acquisition pressure.

[0036] In one embodiment, in step S1, step S12, performing target detection and recognition on the SAR image to obtain a target detection and recognition accuracy rate includes: Step S121, dividing SAR images of different image qualities into a training set and a test set; Step S122: training and testing the target detection and recognition model, and describing the target detection and recognition performance with the target detection and recognition accuracy.

[0037] In one embodiment, the target detection and recognition model is a YOLO-V8 model based on a residual network.

[0038] The YOLO_V8 model is an efficient single-stage object detection method. Building on the YOLO_V5 model, it introduces and improves upon state-of-the-art technologies from other YOLO versions, further enhancing performance and flexibility. Its architecture is similar to the YOLO_V5 model, consisting of three main components: the backbone, neck, and head. The backbone and neck networks form the core structure of the YOLO_V8 model. The backbone performs feature extraction, the neck performs feature fusion, and the head performs prediction. It is widely used in computer vision tasks such as object detection, instance segmentation, and image classification. YOLO_V8 has five models, of which YOLO_V8n offers a lower network depth, smaller size, and faster detection speed, ensuring faster detection speed while maintaining accuracy.

[0039] Specifically, the training set and test set are divided into 8:2 ratios; the deep network is initialized using the KAIMING initialization method, and the image input size is set to 512. 512 1. The batch size is set to 32, and the stochastic gradient descent optimizer is used. The intersection-over-union ratio threshold is set to 0.5, and the number of training rounds is set to 300. The loss changes during training are as follows: Figure 3 As shown in the figure, after the loss no longer decreases, the accuracy remains stable and the model converges. At the end of the training, the YOLO_V8 network model for detection and recognition of the FUSAR-ship dataset is obtained.

[0040] 3600 SAR images for SAR image quality evaluation were selected and input into the trained YOLO_V8 network model. The target detection and recognition results of each image were output, that is, the target detection and recognition accuracy was used to describe the target detection and recognition performance.

[0041] In one embodiment, in step S2, the image blur evaluation index is a blur evaluation index based on the image histogram. BAH (Blur Assessment of Histogram, BAH ).

[0042] In one embodiment, step S2, calculating the blur evaluation index based on the image histogram BAH include: Step S21: Calculate the grayscale histogram of the SAR image and calculate the distribution probability of each grayscale level. ; The calculation formula is: ; in, The gray value in the SAR image is equal to x The number of pixels, is the total number of pixels in the SAR image; Step S22: Calculate the weight of each grayscale value , the weight reflects the distance from the image grayscale mean; The calculation formula is: ; in, is the grayscale mean of the image; Step S23: Calculate the degree evaluation index BAH ; The calculation formula is: .

[0043] Through the above scheme, the image histogram-based ambiguity evaluation index BAH is obtained. It only relies on the histogram statistical characteristics of the SAR image to be evaluated and does not require clear original image comparison. It is applicable to real scenes and avoids situations where there is no reference image or the reference image is difficult to obtain. At the same time, the image histogram-based ambiguity evaluation index BAH is more adapted to SAR characteristics and can avoid local noise interference through global statistics. Even in low contrast (such as the sea surface), the histogram features can still effectively reflect the degree of blur. Finally, the histogram statistics and feature extraction have low computational complexity and high computational efficiency.

[0044] In one embodiment, in step S3, the image noise level evaluation index includes a noise level evaluation index based on the local standard deviation. LS (Signal to noise ratio based on local standard deviation, LSD-SNR, referred to as LS), noise level evaluation index based on wavelet transform WN (Wavelet Noise, abbreviated as WN).

[0045] In one embodiment, step S31, calculate the noise level evaluation index based on the local standard deviation LS include: Step S311: Divide a SAR image into several pixel-sized image blocks and calculate the local mean of each image block. E and local standard deviation S； The calculation formula is: , ; in, Indicates the i The total number of pixels in the image block, Indicates the i In the image block j The grayscale value of a pixel.

[0046] Specifically, if a SAR image is divided into several pixels of size 4 4 or 8 8 image blocks.

[0047] Step S312: Divide the area between the maximum local standard deviation and the minimum local standard deviation by a set multiple into several equally spaced intervals, the local standard deviations of all image blocks fall into the corresponding intervals, and count the number of image blocks in each interval.

[0048] Specifically, when the size of the SAR image is larger than 256 When the value is 256, the area between the maximum local standard deviation and 1.5 times the minimum local standard deviation is divided into 100 equally spaced intervals. The local standard deviations of all image blocks are made to fall into the corresponding intervals. The number of image blocks in each interval is counted. The noise value of the SAR image is the average local standard deviation in the interval with the largest number of blocks.

[0049] Step S313: Calculate the signal-to-noise ratio to obtain LS ; The calculation formula is: ; in, N is the total number of image blocks, M is the number of image blocks in the interval with the largest number of image blocks, arrive q Number the image blocks that fall into the above interval.

[0050] In one embodiment, step S32, calculating the noise level evaluation index based on wavelet transform WN include: The calculation formula is: ; in, Refers to the set frequency information in the diagonal direction of the image.

[0051] Through the above scheme, a noise level evaluation index (LS) based on local standard deviation and a noise level evaluation index (WN) based on wavelet transform are obtained respectively. They can complement each other when used together to evaluate SAR image quality, and can comprehensively and efficiently evaluate the noise characteristics of SAR images. For example, they can cover a full range of noise types and adapt to the needs of different scenarios. LS excels at evaluating random noise in the spatial domain (such as SAR coherent speckle and Gaussian noise) and quantifies noise intensity through local standard deviation. WN excels at capturing periodic noise in the frequency domain (such as scan line streaks and sensor fixed pattern noise) and separates noise through high-frequency sub-band energy. The combination of the two can cover the vast majority of noise types (random + periodic) in SAR images, comprehensively evaluate SAR image noise from both spatial and frequency domain dimensions, perform multi-dimensional quality evaluation, avoid the limitations of a single index, and provide a more reliable noise evaluation benchmark for SAR image evaluation.

[0052] In one embodiment, step S4, establishing a correlation model between the image blur evaluation index, the image noise level evaluation index, and the target detection and recognition accuracy based on a linear regression method to obtain SAR image quality parameters for evaluating SAR image quality, includes: Step S41: Using the image blur evaluation index and the image noise level evaluation index as predictor variables and the target detection and recognition accuracy as the response variable, linear regression fitting is performed to obtain a linear regression model under the FUSAR-ship dataset; The linear regression model is expressed as: ; ; in, BAH is the image blur evaluation index, LS and WN is the image noise level evaluation index, P The target detection and recognition accuracy; A 1. A 2. A 3 and B is the linear regression coefficient.

[0053] Step S42: Generate SAR image quality parameters using the linear regression coefficients of the linear regression model. OQ (Object Detection and Recognition–Quality Evaluation, ODR-QE, referred to as OQ), expressed as: .

[0054] Through the above scheme, the image blur evaluation index reflects the degree of image detail retention, and the image noise level evaluation index reflects the signal interference intensity. Both are quantifiable physical parameters in the SAR imaging process. Image blur (such as resolution degradation) and noise level (such as Gaussian noise), the two physical indicators that most directly affect SAR image quality, are used as predictive variables. A linear mapping relationship is established with the target detection and recognition accuracy (the core requirement of the application end). The application performance is directly associated through a linear regression model. The OQ formed by the weighted combination of the coefficients of the linear regression model intuitively reflects the contribution weight of blur and noise to the accuracy, unifies the evaluation standards, and is both physically interpretable and task-oriented. It is particularly suitable for SAR images in scenarios with strict detection accuracy requirements such as military reconnaissance and disaster monitoring. It can also be extended to the quality evaluation of visible light and infrared images.

[0055] Specifically, 35 typical SAR images with different image qualities were selected from a SAR image set containing 3600 SAR images, and the image fuzziness evaluation index was calculated according to the method provided in this application. BAH , Image noise level evaluation index LS and WN Get SAR image quality parameters OQ, Predicting target detection and recognition accuracy through linear regression model and the target detection recognition accuracy of the target detection model P For comparison, the results of 5 SAR images are shown in Table 1, and the results of 35 SAR images are shown in Figure 4 shown.

[0056]

[0057] As shown in Table 1 and Figure 4 As shown, the target detection and recognition accuracy is predicted by the linear regression model provided by this application Target detection recognition accuracy compared to target detection models P The results are similar and can accurately predict the image detection and recognition situation, proving the effectiveness of the image quality evaluation method provided in this application.

[0058] According to the method provided in this application, the configuration of the SAR system platform can be optimized according to the specific detection and recognition task, so that the SAR image obtained has the image parameters required for the detection and recognition task. On the other hand, for the acquired SAR image, its detection and recognition performance can be predicted, providing guidance for the screening of effective images and reducing labor costs. For example, when the image set is used for recognition network training, SAR images with image parameters that do not meet the standards can be excluded, so that the trained recognition network has better results.

[0059] It should be noted that the above-mentioned serial numbers of the embodiments of the present application are only for description and do not represent the advantages and disadvantages of the embodiments. The terms "including" and "having" in the specification and claims of this application and the above-mentioned drawings, as well as any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device comprising a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units inherent to these processes, methods, products or devices. The terms "first", "second" and "third" are used to distinguish different objects, etc., and do not represent a sequential order, nor do they limit "first", "second" and "third" to being different types.

[0060] In the description of the embodiments of this application, the words "exemplary," "for example," or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary," "for example," or "for example" in the embodiments of this application should not be construed as being preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary," "for example," or "for example" is intended to present the relevant concepts in a concrete manner.

[0061] In the description of the embodiments of the present application, unless otherwise specified, “ / ” means or, for example, A / B can mean A or B; “and / or” in the text is merely a description of the association relationship of associated objects, indicating that three relationships may exist, for example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, in the description of the embodiments of the present application, “multiple” refers to two or more than two.

[0062] In some processes described in the embodiments of this application, multiple operations or steps are included that appear in a specific order. However, it should be understood that these operations or steps may not be executed in the order in which they appear in the embodiments of this application or may be executed in parallel. The sequence numbers of the operations are only used to distinguish different operations, and the sequence numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed in sequence or in parallel, and these operations or steps may be combined.

[0063] The above are only preferred embodiments of the present application and do not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A SAR image quality assessment method based on target detection and recognition, characterized in that: It includes the following steps: Acquire a SAR image set comprising a plurality of SAR images, and perform target detection and recognition on the SAR images to obtain a target detection and recognition accuracy rate; Calculating an image fuzziness evaluation index of the SAR image; Calculating an image noise level evaluation index of the SAR image; Based on the linear regression method, a correlation model between the image blur evaluation index, the image noise level evaluation index and the target detection and recognition accuracy is established to obtain a SAR image quality parameter for evaluating the SAR image quality.

2. The SAR image quality assessment method based on target detection and recognition according to claim 1, wherein: The linear regression method is used to establish a correlation model between the image blur evaluation index, the image noise level evaluation index and the target detection and recognition accuracy to obtain the SAR image quality parameters for evaluating the SAR image quality, including: Using the image blur evaluation index and the image noise level evaluation index as predictor variables and the target detection and recognition accuracy rate as the response variable, linear regression fitting is performed to obtain a linear regression model; The linear regression model is expressed as: ； ； in, BAH is the image blur evaluation index, LS and WN is the image noise level evaluation index, P The target detection and recognition accuracy; The SAR image quality parameters are formed by the linear regression model OQ , expressed as: 。 3. The SAR image quality assessment method based on target detection and recognition according to claim 2, wherein: The image fuzziness evaluation index is a fuzziness evaluation index based on the image histogram BAH .

4. The SAR image quality assessment method based on target detection and recognition according to claim 3, wherein: Calculate the image histogram-based fuzziness evaluation index BAH include: Statistically calculate the grayscale histogram of the SAR image and the distribution probability of each grayscale level ; The calculation formula is: ; in, The gray value in the SAR image is equal to x The number of pixels, is the total number of pixels in the SAR image; Calculate the weight of each grayscale value ; The calculation formula is: ; in, is the grayscale mean of the image; Calculate the degree evaluation index BAH ; The calculation formula is: .

5. The SAR image quality assessment method based on target detection and recognition according to claim 2, wherein: The image noise level evaluation index includes a noise level evaluation index based on local standard deviation LS , Noise level evaluation index based on wavelet transform WN .

6. The SAR image quality assessment method based on target detection and recognition according to claim 5, wherein: Calculate the noise level evaluation index based on local standard deviation LS include: Divide a SAR image into several pixel-sized image blocks and calculate the local mean of each image block. E and local standard deviation S； The calculation formula is: , ; in, Indicates the i The total number of pixels in the image block, Indicates the i In the image block j The grayscale value of each pixel; Divide the area between the maximum local standard deviation and the minimum local standard deviation by a set multiple into a number of equally spaced intervals, the local standard deviations of all the image blocks fall into the corresponding intervals, and count the number of image blocks in each interval; Calculate the signal-to-noise ratio to get LS ; The calculation formula is: ; in, N is the total number of image blocks, M is the number of image blocks in the interval with the largest number of image blocks, p arrive q Number the image blocks that fall into the above interval.

7. The SAR image quality assessment method based on target detection and recognition according to claim 5, wherein: Calculate the noise level evaluation index based on wavelet transform WN include: The calculation formula is: ; in, Refers to the set frequency information in the diagonal direction of the image.

8. The SAR image quality assessment method based on target detection and recognition according to claim 1, wherein: The acquiring of a SAR image set comprising a plurality of SAR images comprises: Get SAR original image; increasing the image blur and noise level of the SAR original image to obtain a plurality of degraded images; The SAR original image and the degraded image are used as the SAR images to form the SAR image set.

9. The SAR image quality assessment method based on target detection and recognition according to claim 8, wherein: The performing target detection and recognition on the SAR image to obtain a target detection and recognition accuracy rate includes: Dividing the SAR images of different image qualities into a training set and a test set; The target detection and recognition model is trained and tested, and the target detection and recognition performance is described by the target detection and recognition accuracy.

10. The SAR image quality assessment method based on target detection and recognition according to claim 9, wherein: The target detection and recognition model is a YOLO-V8 model based on a residual network.

Images