Motion blur detection method and system based on deep learning and blind deconvolution

Abstract

The invention relates to the technical field of blurring detection, and discloses a motion blurring detection method and system based on deep learning and blind deconvolution, and the method comprises the steps: collecting an original image, and carrying out the sub-region division of the image; singular value decomposition is carried out on each sub-region; performing fuzzy region detection on the sub-regions according to a decomposition result; performing image restoration on the blurred image; and segmenting target contour information in the fuzzy restored image by using a deep learning method, and outputting a clear image with a target contour. By introducing the SPD-Conv, the improved YOLOv8 model can more accurately segment a low-resolution fuzzy target, and the segmentation precision is effectively improved. And the method can adapt to low-resolution fuzzy target segmentation tasks in different scenes. While the segmentation precision is improved, the calculation complexity and the reasoning time of the model are not remarkably increased, the real-time advantage of the YOLOv8 can still be kept, and the requirement for real-time performance in practical application is met.

Description

Technical Field

[0001] The present invention relates to the field of blur detection technology, and specifically to a motion blur detection method and system based on deep learning and blind deconvolution. Background Art

[0002] As the global demand for clean energy continues to grow, wind power generation has been widely developed as a renewable energy technology. As the core equipment of wind power generation, the safe operation of wind turbines is of vital importance. In the wind turbine wind power clearance detection project, accurately detecting the status of wind turbine blades is one of the key links to ensure the safe operation of wind turbines.

[0003] In order to achieve all-weather wind turbine clearance safety monitoring, it is necessary to segment the outline of the wind turbine blades in real time in a dark environment. In order to obtain a relatively clear image, it is usually necessary to extend the camera exposure time to obtain the target features. Due to the high rotation speed of the wind turbine blades, it is inevitable that the wind turbine blades will have serious motion blur in the imaging. This blur has a great impact on the image quality, making the outline of the wind turbine blades unclear and greatly reducing the segmentation accuracy.

[0004] The existing related technologies for removing motion blur commonly include inverse filtering, Wiener filtering, LR algorithm and some deep neural networks, such as DeblurGAN. Inverse filtering and Wiener filtering data non-blind deconvolution require obtaining the blur kernel of the image, and the blur restoration effect of images with large noise is not good. The LR algorithm needs to be continuously iterated to remove image blur. If the original image is guaranteed to be large, the final image restoration quality will be extremely poor. DeblurGan, as a newcomer in the field of deblurring in deep learning methods, has a high quality of blurry image restoration, but the method is difficult to reproduce. Summary of the invention

[0005] In view of the above-mentioned problems, the present invention is proposed.

[0006] Therefore, the technical problem solved by the present invention is that the existing blurred image restoration method has the problems of great difficulty in reproduction and inability to accurately segment the moving part.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions: a motion blur detection method based on deep learning and blind deconvolution, comprising:

[0008] Collect the original image and divide the image into sub-regions;

[0009] Perform singular value decomposition on each sub-region separately;

[0010] According to the decomposition result, performing fuzzy area detection on the sub-area;

[0011] Perform image restoration on blurred images;

[0012] Use deep learning methods to segment target contour information in the fuzzy restored image and output a clear image with target contour.

[0013] As a preferred solution of the motion blur detection method based on deep learning and blind deconvolution described in the present invention, the sub-region division includes dividing the original input image into sub-regions of fixed sizes.

[0014] As a preferred solution of the motion blur detection method based on deep learning and blind deconvolution described in the present invention, the singular value decomposition includes: assuming that the image I is an m×n image, then U is an m×m orthogonal matrix, whose column vectors are left singular vectors; V is an n×n orthogonal matrix, whose column vectors are right singular vectors; S is an m×n diagonal matrix, and the elements on the diagonal are are non-negative real numbers and are in descending order of singular values;

[0015] For the singular value decomposition of the blurred image I, it is expressed as:

[0016] ;

[0017] Wherein, i represents the diagonal index, i∈{1, 2, ..., min(m, n)}; T represents transpose.

[0018] As a preferred solution of the motion blur detection method based on deep learning and blind deconvolution described in the present invention, the blurred area detection includes: for each sub-area Do singular value decomposition and extract singular values ; Sort the singular value information in descending order, and assume that the singular values ​​of the main feature information in the corresponding image are k, ;

[0019] Calculate the sum of the ratios of each value in the singular value information to the initial value:

[0020] ;

[0021] in, is the image area The quality evaluation score of ; j represents the index of the diagonal line in the diagonal matrix of the sub-region; Represents decomposed image After the feature information of the subject is obtained, the first k feature values ​​are the singular values ​​of the subject texture feature; Represents the key information part of the image and the corresponding singular value;

[0022] For image area Apply Gaussian filtering blur, generate reference image, and calculate quality evaluation score ;

[0023] The change in the singular values ​​of the image is:

[0024] ;

[0025] When diff is greater than the threshold , then determine the image sub-region Belongs to the fuzzy area.

[0026] As a preferred solution of the motion blur detection method based on deep learning and blind deconvolution described in the present invention, the image restoration includes: assuming that the image is formed by convolving a clear image with a blur kernel and adding noise;

[0027]

[0028] Among them, g represents blurred image, h represents high-definition image, f(x) represents blur kernel, and n(x) represents noise;

[0029] The blind deconvolution method based on MLE obtains the optimal combination of clear image and blur kernel by constructing the likelihood function. Reach the maximum value;

[0030] Assume that the noise n(x) has a mean of 0 and a variance of Gaussian distribution, the blurred image is observed The probability density function of is:

[0031]

[0032] The goal of MLE is to maximize the likelihood function, and the log-likelihood function is obtained:

[0033]

[0034] Where C is a constant, the equation is equivalent to the least squares minimization .

[0035] As a preferred solution of the motion blur detection method based on deep learning and blind deconvolution described in the present invention, the specific steps of restoring the blurred image are:

[0036] Select initial values ​​to guess blur kernel f(x) and clear image g;

[0037] Fix h and update f(x): Let h be fixed as , then solve ;

[0038] Determine whether the algorithm converges. If the convergence condition is not met, continue to iterate; otherwise, output the estimated clear image h and blur kernel f(x).

[0039] As a preferred solution of the motion blur detection method based on deep learning and blind deconvolution described in the present invention, wherein: the method of using deep learning to segment the target contour information in the blurred restored image includes: using the improved yolov8 to identify the target contour of the restored image and output a feature map;

[0040] The improved yolov8 includes inserting an SPD module on the basis of yolov8 and replacing the conv-c2f of the backbone with spd-c2f.

[0041] A motion blur detection system based on deep learning and blind deconvolution using any method as described in the present invention, characterized in that:

[0042] An acquisition module acquires the original image and divides the image into sub-areas;

[0043] The detection module performs singular value decomposition on each sub-region respectively; and performs fuzzy region detection on the sub-region according to the decomposition result;

[0044] A processing module is used to restore the blurred image;

[0045] The output module uses deep learning methods to segment the target contour information in the fuzzy restored image and outputs a clear image with the target contour.

[0046] A computer device comprises: a memory and a processor; the memory stores a computer program, wherein: the processor implements the steps of any one of the methods of the present invention when executing the computer program.

[0047] A computer-readable storage medium stores a computer program, wherein: when the computer program is executed by a processor, the steps of any one of the methods of the present invention are implemented.

[0048] Beneficial effects of the present invention: The motion blur detection method based on deep learning and blind deconvolution provided by the present invention, by introducing SPD-Conv, the improved YOLOv8 model can more accurately segment low-resolution blurred targets, effectively improve the segmentation accuracy, and provide a more reliable basis for subsequent target recognition, analysis and processing. Experimental results on a variety of low-resolution blurred image data sets show that the method proposed in the present invention has good generalization ability and can adapt to low-resolution blurred target segmentation tasks in different scenarios. While improving the segmentation accuracy, the method proposed in the present invention does not significantly increase the computational complexity and reasoning time of the model, and can still maintain the real-time advantage of YOLOv8, meeting the real-time requirements in practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0050] Figure 1 An overall flow chart of detecting fan blades in a motion blur detection method based on deep learning and blind deconvolution provided in a first embodiment of the present invention;

[0051] Figure 2 A comparison diagram of blur detection of a motion blur detection method based on deep learning and blind deconvolution provided in the second embodiment of the present invention; wherein (a) represents an image before region detection, and (b) represents an image after region detection;

[0052] Figure 3 A comparison diagram of blurred image restoration of a motion blur detection method based on deep learning and blind deconvolution provided in the second embodiment of the present invention; wherein (a) represents the original image before restoration, and (b) represents the image after restoration;

[0053] Figure 4 A comparison diagram of target segmentation of a motion blur detection method based on deep learning and blind deconvolution provided in the second embodiment of the present invention; wherein (a) represents the image before segmentation, and (b) represents the result after segmentation. DETAILED DESCRIPTION

[0054] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0055] Example 1, reference Figure 1 , is an embodiment of the present invention, and provides a motion blur detection method based on deep learning and blind deconvolution, comprising:

[0056] S1: Collect the original image and divide the image into sub-regions.

[0057] First, the original input image is divided into several small sub-regions (for example, 9×9). This is because the direct singular value decomposition calculation of the original image is computationally intensive and time-consuming, and the blur degree and blur type of each region of the image are different. Dividing the image into several small sub-regions can accelerate the singular value decomposition process of the image data, and can also process data of different blur types and different blur degrees to avoid the loss of blur features.

[0058] S2: Perform singular value decomposition on each sub-region separately.

[0059] It should be said that the blurred image is essentially composed of the original high-definition image and the blur kernel. The blur kernel will smooth the high-frequency features in the image, such as the contour features of the object, and make them disappear. Image data is actually a set of pixel matrices. The contour features in the image can be shown from the numerical distribution characteristics of the pixel matrix. The pixel matrix singular value distribution of a clear image has relatively obvious characteristics. Generally speaking, the decay of its singular values ​​is relatively slow. For example, for a natural landscape image with rich details and clear edges, the first few singular values ​​will be relatively large, because these singular values ​​mainly correspond to the main structures in the image (such as the horizon, mountain contours, etc.) and high-energy parts. As the singular value sequence increases, the singular value will gradually decrease, but it will not approach zero quickly. This is because a clear image contains a variety of frequency components, from low-frequency large-area color areas to high-frequency fine textures and edges, and this information is reflected in the singular value decomposition.

[0060] Blurring will cause the singular value distribution of the image pixel matrix to change. First, the blur operation is similar to a low-pass filter, which will cause the image to lose high-frequency information. In the singular value distribution, it manifests as a rapid decay of the singular value.

[0061] For the singular value decomposition of the blurred image I, it can be expressed as:

[0062]

[0063] Assuming that image I is an m×n image, then U is an m×m orthogonal matrix whose column vectors are left singular vectors; V is an n×n orthogonal matrix whose column vectors are right singular vectors;

[0064]

[0065] S is an m×n diagonal matrix with the elements on the diagonal are non-negative real numbers and are in descending order of singular values, .

[0066] S3: Perform fuzzy area detection on the sub-region according to the decomposition result.

[0067] For each sub-region Perform SVD singular value decomposition and extract its singular value information , among which The singular value information corresponds to the main feature information in the image, such as the contour feature; the rest corresponds to the detail information of the image. Gaussian blur will reduce the image detail features, that is, increase the singular value. Compared with the clear image, the blurred image The main feature information accounts for a large proportion, so the ratio of each value in the singular value information to the initial value and the ratio can be used as the evaluation image blur score:

[0068] ;

[0069] in, is the image area The quality evaluation score of ; j represents the index of the diagonal line in the diagonal matrix of the sub-region; Represents decomposed image After the feature information, sort in descending order, The eigenvalues ​​are singular values ​​of the main texture features, and the rest are detail features;

[0070] According to the singular value decomposition theory, image I can be understood by first After rotation, in the new coordinate system Zoom in, then pass Rotate back, It is in this optimal (in a certain mathematical sense, such as the transformation that minimizes energy loss) transformation process that the maximum degree of scaling in a certain direction carries the most important, holistic key information or the strongest energy component of the image.

[0071] For image area Perform Gaussian filtering blur (Gaussian kernel is 3, normal distribution filling), generate reference image, and calculate quality evaluation score .

[0072] The change in the singular values ​​of the image is:

[0073]

[0074] When diff is greater than the threshold , then the current image sub-region Belongs to the fuzzy area.

[0075] S4: Perform image restoration on the blurred image.

[0076] Assume that the image is formed by convolving the clear image with the blur kernel and adding noise:

[0077]

[0078] Among them, g represents the blurred image, h represents the high-definition image, f(x) represents the blur kernel, and n(x) represents the noise.

[0079] The MLE-based blind deconvolution method constructs a likelihood function to find the most likely combination of a clear image and a blur kernel, so that the likelihood function reaches its maximum value given the observed blurry image. The likelihood function is usually constructed based on the probability distribution assumption of the noise. For example, if the noise is Gaussian distributed, the likelihood function can be constructed based on the probability density function of the Gaussian distribution.

[0080] Assume that the noise n(x) has a mean of 0 and a variance of Gaussian distribution, then the blurred image is observed The probability density function of is:

[0081]

[0082] The goal of MLE is to maximize the likelihood function, and the log-likelihood function is obtained:

[0083]

[0084] Where C is a constant, the equation is equivalent to the least squares minimization .

[0085] The steps of blur image restoration include:

[0086] Choose appropriate initial values ​​to guess the blur kernel f(x) and the clear image g. The initialization parameter value will affect the convergence speed of the algorithm and the final result.

[0087] Fix h and update f(x): Let h be fixed as , then solve .

[0088] Determine whether the algorithm converges. The convergence condition can be that the change of the log-likelihood function is less than a certain threshold, or the relative change of f(x) is less than a certain threshold, etc. If the convergence condition is not met, continue to iterate; otherwise, output the estimated clear image h and blur kernel f(x).

[0089] S5: Use deep learning methods to segment target contour information in the blurred restored image and output a clear image with target contour.

[0090] Use deep learning methods to segment target contour information in fuzzy restored images, insert the SPD module based on yolov8 (replace the backbone's conv-c2f with spd-c2f), introduce global context and local feature information, improve the model's feature extraction capabilities in low resolution and fine contours, and improve its robustness in complex scenes.

[0091] The meaning of each layer of SPD module:

[0092] (a) shows the standard feature map.

[0093] (b) shows the spatial-to-depth operation, where the spatial information is rearranged into the depth channel.

[0094] (c) shows the increased depth of the resulting feature map.

[0095] (d) Represents a non-strided convolutional layer applied after the SPD operation.

[0096] (e) shows the output feature map after a convolution with a stride of 1, which maintains the spatial resolution but changes the depth dimension.

[0097] On the other hand, this embodiment also provides a motion blur detection system based on deep learning and blind deconvolution, which includes:

[0098] The acquisition module acquires the original image and divides the image into sub-areas.

[0099] The detection module performs singular value decomposition on each sub-region respectively; and performs fuzzy region detection on the sub-region according to the decomposition result.

[0100] The processing module restores the blurred image.

[0101] The output module uses deep learning methods to segment the target contour information in the fuzzy restored image and outputs a clear image with the target contour.

[0102] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program code.

[0103] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0104] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0105] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0106] Embodiment 2, as Figure 2-Figure 4 As an embodiment of the present invention, an implementation process of a motion blur detection method based on deep learning and blind deconvolution is provided.

[0107] (1) Data preprocessing: Preprocess the input low-resolution blurred image, including image normalization, cropping, scaling, etc., to make it meet the requirements of network input.

[0108] (2) Model training: The improved YOLOv8 model is trained using a low-resolution blurred target image dataset with annotations. During the training process, the model parameters, such as learning rate and weight decay, are adjusted to optimize the performance of the model. At the same time, data enhancement techniques, such as rotation, flipping, and adding noise, can be used to increase the diversity of the dataset and improve the generalization ability of the model.

[0109] (3) Model reasoning: The low-resolution blurred image to be segmented is input into the trained model, and after forward propagation calculation, the target segmentation result is obtained. During the reasoning process, the output result can be post-processed according to actual needs, such as removing noise, filling holes, etc., to obtain a more accurate and smoother segmentation mask.

[0110] (4) Performance evaluation: The segmentation performance of the model is evaluated using indicators such as accuracy, recall, and mean average precision (mAP). The effectiveness of the proposed method in improving the accuracy of low-resolution blurred target segmentation is verified by comparative experiments with the original YOLOv8 model and other existing segmentation methods.

[0111] After the singular value decomposition of the blurred image I, the blurred area detection is performed to obtain the blurred area, such as Figure 2 As shown, the left side represents the image before region detection; the right side represents the image after region detection, where the red part represents the detected background part, and the other parts are the detected blurred areas.

[0112] After the detection is completed, the blurred part is restored, such as Figure 3As shown, the left side shows the original image before restoration, and the right side shows the restored image. It can be seen that the present invention can effectively clarify the blurred image.

[0113] Perform target segmentation on the cleared image, such as Figure 4 As shown, the left side represents the image before segmentation, the right side represents the image after segmentation, and red represents the target object, thus realizing blurred detection of moving objects.

[0114] The core of the present invention is to introduce SPD-Conv into the YOLOv8 network structure, and process the input feature map through SPD-Conv, which effectively improves the model's feature extraction ability for low-resolution blurred targets, thereby improving the segmentation accuracy.

[0115] SPD-Conv principle and advantages: SPD-Conv consists of a space-to-depth (SPD) layer and a non-stride convolution (Conv) layer. The SPD layer reduces each spatial dimension of the input feature map to the channel dimension while retaining the information within the channel, reducing the size of the spatial dimension; the Conv layer performs a standard convolution operation after the SPD layer, avoiding the oversampling problem that may be caused by stride convolution and retaining more fine-grained information. This combination can improve the model's detection performance for low-resolution images and small objects, and reduce dependence on high-quality input.

[0116] Improved YOLOv8 network structure: In the backbone network and / or neck network of YOLOv8, some traditional convolutional layers are replaced by SPD-Conv modules. Specifically, according to the characteristics and requirements of the target segmentation task, the SPD-Conv module can be inserted in a suitable position to give full play to its advantages and enhance the network's ability to capture the features of low-resolution blurred targets.

[0117] Principle of improved segmentation accuracy: The feature information of low-resolution blurred targets is relatively weak and difficult to extract, and traditional convolutional layers may lose important information when processing such targets. SPD-Conv can better retain and utilize this feature information, and by converting spatial information into depth information, the network can more comprehensively learn the feature representation of the target, thereby more accurately segmenting low-resolution blurred targets.

[0118] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A motion blur detection method based on deep learning and blind deconvolution, characterized in that: include: Collect the original image and divide the image into sub-regions; Perform singular value decomposition on each sub-region separately; According to the decomposition result, performing fuzzy area detection on the sub-area; Perform image restoration on blurred images; Use deep learning methods to segment target contour information in the fuzzy restored image and output a clear image with target contour; The image restoration includes: assuming that the image is formed by convolving a clear image with a blur kernel and adding noise, and forming a clear image through iteration; ， Among them, g represents blurred image, h represents high-definition image, f(x) represents blur kernel, and n(x) represents noise; The blind deconvolution method based on MLE obtains the optimal combination of clear image and blur kernel by constructing the likelihood function. Reach the maximum value; Assume that the noise n(x) has a mean of 0 and a variance of Gaussian distribution, the blurred image is observed The probability density function of is: ， The goal of MLE is to maximize the likelihood function, and the log-likelihood function is obtained: ， Where C is a constant, the equation is equivalent to the least squares minimization .

2. The motion blur detection method based on deep learning and blind deconvolution as claimed in claim 1, characterized in that: The sub-region division includes dividing the original input image into sub-regions of fixed size.

3. The motion blur detection method based on deep learning and blind deconvolution as claimed in claim 2, characterized in that: The singular value decomposition includes: assuming that the image I is an m×n image, then U is an m×m orthogonal matrix, whose column vectors are left singular vectors; V is an n×n orthogonal matrix, whose column vectors are right singular vectors; S is an m×n diagonal matrix, and the elements on the diagonal are are non-negative real numbers and are in descending order of singular values; The singular value decomposition of the blurred image I is expressed as: ； Wherein, i represents the diagonal index, i∈{1, 2, ..., min(m, n)}; T represents transpose.

4. The motion blur detection method based on deep learning and blind deconvolution as claimed in claim 3, characterized in that: The fuzzy area detection includes: for each sub-area Do singular value decomposition and extract singular values ; Sort the singular value information in descending order, and assume that the singular values ​​of the main feature information in the corresponding image are k, ; Calculate the sum of the ratios of each value in the singular value information to the initial value: ； in, is the image area The quality evaluation score of ; j represents the index of the diagonal line in the diagonal matrix of the sub-region; Represents decomposed image After the feature information of the subject is obtained, the first k feature values ​​are the singular values ​​of the subject texture feature; Represents the key information part of the image and the corresponding singular value; For image area Use Gaussian filter blur to generate reference image and calculate quality evaluation score ; The change in the singular values ​​of the image is: ； When diff is greater than the threshold , then determine the image sub-region Belongs to the fuzzy area.

5. The motion blur detection method based on deep learning and blind deconvolution as claimed in claim 4, characterized in that: The specific steps of restoring the blurred image are as follows: Select initial values ​​to guess blur kernel f(x) and clear image g; Fix h and update f(x): Let h be fixed as , then solve ; Determine whether the algorithm converges. If the convergence condition is not met, continue to iterate; otherwise, output the estimated clear image h and blur kernel f(x).

6. The motion blur detection method based on deep learning and blind deconvolution as claimed in claim 5, characterized in that: The method of using deep learning to segment target contour information in the fuzzy restored image includes: using the improved yolov8 to identify the target contour of the restored image and output a feature map; The improved yolov8 includes inserting an SPD module on the basis of yolov8 and replacing the conv-c2f of the backbone with spd-c2f.

7. A motion blur detection system based on deep learning and blind deconvolution using the method according to any one of claims 1 to 6, characterized in that: An acquisition module acquires the original image and divides the image into sub-areas; The detection module performs singular value decomposition on each sub-region; According to the decomposition result, performing fuzzy area detection on the sub-area; A processing module is used to restore the blurred image; The output module uses deep learning methods to segment the target contour information in the fuzzy restored image and outputs a clear image with the target contour.

8. A computer device comprising: Memory and processor; The memory stores a computer program, wherein the processor implements the steps of any one of the methods of claims 1-6 when executing the computer program.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images