A method for detecting blur in ultraviolet images based on integer approximate factorization of DFT matrix

Abstract

The application discloses a UV image blur detection method based on DFT matrix integer approximate decomposition, which performs integer approximate decomposition on a matrix on the basis of setting a numerical range constraint and a sparsity constraint when the DFT matrix is used for image discrete Fourier transform. The matrix is decomposed into a form of a series of sparse and element value limited matrix multiplication, so as to reduce the matrix multiplication calculation times, thereby optimizing the calculation complexity and reducing the calculation cost. The spectrum image obtained by performing the discrete Fourier transform on the UV image of the water surface floating colorless chemical based on the method is subjected to high-pass filtering, the average amplitude intensity is calculated, the blur value is compared with a set threshold value, whether the UV image of the water surface floating colorless chemical is blurred can be determined, and thus the blur detection function is realized.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of chemical leakage monitoring, and particularly relates to a UV image blur detection method based on DFT matrix integer approximate decomposition. BACKGROUND

[0002] Marine chemical leakage accidents occur frequently, causing serious casualties, property losses and marine environmental pollution, and having a disastrous impact on the marine economy and marine environment. The UV sensor has strong detection capability for thin liquid films, but when the on-board UV imaging equipment is used to shoot the UV image of the leaked floating colorless chemicals, the relative motion between the shot chemical target and the shooting equipment occurs due to the shaking of the shooting equipment and the influence of the sea waves, resulting in blurred UV image. In the blurred UV image, the boundary of the chemicals is not clear and complete, which greatly affects the subsequent chemical target segmentation task, so that the chemicals target cannot be accurately extracted from the water background, and the area and distribution of the leakage area cannot be well determined.

[0003] Therefore, the blur detection method for detecting whether the UV image shot at the chemical leakage accident site is blurred can provide a blurred image for subsequent image deblurring processing, avoiding manual screening of blurred images, and having great significance for chemical leakage processing. SUMMARY

[0004] In view of the deficiencies of the prior art, the purpose of the present application is to provide a new method for judging whether the UV image of the floating colorless chemicals on the water surface is blurred, that is, using the method based on DFT matrix integer approximate decomposition to realize the UV image blur detection of the floating colorless chemicals on the water surface. The present application can automatically screen whether the shot chemical UV image is blurred, avoiding manual screening and improving the efficiency of screening the blurred chemical UV image.

[0005] The present application is realized by the following technical solutions:

[0006] A UV image blur detection method based on DFT matrix integer approximate decomposition, comprising the following steps:

[0007] (1) The UV camera samples different shot targets to obtain the UV images corresponding to different targets;

[0008] (2) multiplying the two-dimensional matrix corresponding to the chemical UV image with the DFT matrix to perform discrete Fourier transform on the UV image to obtain a frequency spectrum image;

[0009] (3) shielding the center of the frequency spectrum image after the centering processing and the low-frequency information of the image to complete the high-pass filtering of the frequency spectrum image;

[0010] (4) calculating the average high frequency amplitude of the high-pass filtered frequency spectrum image as the blur value of the image for blur judgment;

[0011] (5) setting a threshold value of image definition, comparing the blur value of the image with the threshold value, and determining the image as a blurred image when the blur value of the image is less than the threshold value.

[0012] Further, the target of the step (1) is a colorless chemical floating on the water surface, and the ultraviolet image spectral range is 10nm-400nm.

[0013] Further, in the step (2), the DFT matrix is obtained by integer approximate decomposition, and the matrix size is consistent with the size of the ultraviolet image.

[0014] Further, the integer approximate decomposition is specifically: the DFT matrix is decomposed into the form of multiplication of three sparse matrices, and the element values in the sparse matrices are all integers; and the matrix representation is as follows:

[0015] F N ≈A1A2A3;

[0016] In the formula, F N is the DFT matrix, and A1, A2 and A3 are three sparse matrices obtained by integer approximate decomposition.

[0017] Further, in the step (2), the ultraviolet image is subjected to discrete Fourier transform to obtain a frequency spectrum image, which is specifically: the ultraviolet image of the chemical is multiplied by the three sparse matrices obtained by integer approximate decomposition, and subjected to discrete Fourier transform to obtain the frequency spectrum image of the ultraviolet image of the chemical; and the matrix representation is as follows:

[0018] X=A1A2A3x;

[0019] In the formula, x is the matrix representation of the ultraviolet image of the chemical, and X is the frequency spectrum image of the ultraviolet image after the discrete Fourier transform.

[0020] Further, in the step (3), the frequency spectrum image after the centering processing is specifically: the zero frequency component of the frequency spectrum image is moved to the center position of the frequency spectrum image, and the low frequency information is shielded by shielding the center region of the frequency spectrum image on the basis of the frequency spectrum image.

[0021] Further, in the step (4), the average high frequency amplitude of the image is obtained by statistically calculating the high frequency amplitude of the whole image on the frequency spectrum image after the low frequency information is shielded and then averaging.

[0022] Specifically, the threshold value in the step (5) is manually set according to the quality of the ultraviolet image.

[0023] The application further provides a UV image blur detection device based on DFT matrix integer approximate decomposition.

[0024] An image acquisition module: an ultraviolet camera samples different shooting targets to obtain corresponding ultraviolet images of different targets.

[0025] A discrete Fourier transform module: a two-dimensional matrix corresponding to the ultraviolet image is multiplied by a DFT matrix to perform discrete Fourier transform on the ultraviolet image to obtain a frequency spectrum image.

[0026] A high-pass filter module: the center of the frequency spectrum image after the centering process and low-frequency information of the image are shielded to complete high-pass filtering of the frequency spectrum image.

[0027] A blur value calculation module: mean value calculation is performed on the frequency spectrum image after high-pass filtering to obtain the average high-frequency amplitude of the image as the blur value for blur judgment of the image.

[0028] A blur detection module: a threshold of image sharpness is set, the blur value of the image is compared with the threshold, and when the blur value of the image is less than the threshold, the image is determined to be a blurred image.

[0029] The application has the following beneficial effects:

[0030] The application discloses a UV image blur detection method for water-surface floating colorless chemicals based on DFT matrix integer approximate decomposition.

[0031] The application discloses a UV image blur detection method for water-surface floating colorless chemicals based on DFT matrix integer approximate decomposition.

[0032] The application discloses a UV image blur detection method for water-surface floating colorless chemicals based on DFT matrix integer approximate decomposition. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 The application discloses a UV image blur detection method for water-surface floating colorless chemicals based on DFT matrix integer approximate decomposition. DETAILED DESCRIPTION

[0034] The technical implementation steps involved in the present application will be described in detail and completely below in conjunction with the accompanying drawings in the embodiments of the present application, and the described embodiments are only a part of the embodiments involved in the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of the present application.

[0035] The application discloses a method for detecting the blur of the ultraviolet image of the colorless chemical floating on the water surface based on the integer approximate decomposition of the DFT matrix, which comprises the following steps: taking the ultraviolet image of the chemical floating on the water surface by using an ultraviolet camera; performing the integer approximate decomposition on the DFT matrix corresponding to the size of the ultraviolet image, and converting the DFT matrix into the form of multiplication of three sparse matrices, and the element values of the sparse matrices are integers; multiplying the taken ultraviolet image with the three sparse matrices to perform the discrete Fourier transform, and obtaining the corresponding spectrum image; performing the centering operation on the spectrum image, which is to move the zero frequency component (i.e. the direct current component) of the spectrum image to the center position of the spectrum image, i.e. to concentrate the low frequency information of the image in the center region of the spectrum image and disperse the high frequency information around the spectrum image; shielding the center region of the spectrum image to filter out the low frequency information and realize the high-pass filtering of the centered spectrum image; calculating the average amplitude intensity of the high-pass filtered spectrum image as the blur value of the image; comparing the blur value with the threshold value for judging whether the blur is artificial, if the blur value is lower than the threshold value, the ultraviolet image is judged as a blurred image, otherwise, if the blur value is higher than the threshold value, the ultraviolet image is judged as a clear image; the present application can judge whether the ultraviolet image of the chemical floating on the water surface taken by the ultraviolet camera is blurred, which avoids the artificial visual judgment of whether the taken image is blurred and improves the efficiency of screening the ultraviolet image of the high-quality chemical. Screening the clear ultraviolet image of the chemical is of great significance for the emergency monitoring of the sudden leakage accident of the chemical on the sea.

[0036] Figure 1 The flow chart of the method for detecting the blur of the ultraviolet image of the colorless chemical floating on the water surface based on the integer approximate decomposition of the DFT matrix is shown in the present application.

[0037] As shown in the figure, the method for detecting the blur of the ultraviolet image of the colorless chemical floating on the water surface in the present application comprises the following steps:

[0038] In step S101, the ultraviolet image of the chemical floating on the water surface is taken by using an ultraviolet camera.

[0039] In step S102, the DFT matrix corresponding to the size of the ultraviolet image of the chemical is subjected to the integer approximate decomposition, and the DFT matrix is converted into the form of multiplication of three sparse matrices, and the element values in the sparse matrices are integers. The matrix representation is as follows:

[0040] F N ≈A1A2A3;

[0041] In the formula, F N is a DFT matrix, and A1, A2 and A3 are three sparse matrices obtained through integer approximate decomposition.

[0042] In step S103, the chemical UV image is multiplied by the three sparse matrices obtained through integer approximate decomposition, and discrete Fourier transform is performed to obtain a frequency spectrum of the chemical UV image.

[0043] X=A1A2A3x;

[0044] In the formula, x is a matrix representation of the chemical UV image, and X is a frequency spectrum of the UV image after discrete Fourier transform.

[0045] In step S104, the frequency spectrum is subjected to centering operation, and the specific operation is as follows: the zero frequency component (i.e. the direct current component) of the frequency spectrum is moved to the center position of the frequency spectrum, that is, the low frequency information of the image is concentrated in the center region, and the high frequency information of the image is dispersed around.

[0046] In step S105, high-pass filtering is performed on the frequency spectrum, and this is achieved by shielding the center region of the frequency spectrum.

[0047] In step S106, the average amplitude intensity of the image after high-pass filtering is calculated, and the average amplitude intensity is taken as a blur value.

[0048] In step S107, the blur value is compared with a threshold value of image sharpness given by a human being, if the blur value is higher than the threshold value, it is determined that the UV image is a sharp image, otherwise, if the blur value is lower than the threshold value, it is determined that the UV image is a blurred image.

[0049] The application further provides an UV image blur detection device based on integer approximate decomposition of a DFT matrix, and the device comprises the following modules.

[0050] An image acquisition module: an UV camera samples different shooting targets to obtain UV images corresponding to different targets;

[0051] A discrete Fourier transform module: a two-dimensional matrix corresponding to the UV image is multiplied by a DFT matrix to perform discrete Fourier transform on the UV image to obtain a frequency spectrum image;

[0052] A high-pass filtering module: the center of the frequency spectrum image subjected to centering operation and the low frequency information of the image are shielded to complete high-pass filtering of the frequency spectrum image;

[0053] The blur value calculation module calculates the mean value of the high-pass filtered frequency spectrum image, and obtains the average high-frequency amplitude value of the image as the blur value of the image for blur judgment.

[0054] The blur detection module sets a threshold value of image definition, compares the blur value of the image with the threshold value, and determines that the image is a blurred image when the blur value of the image is less than the threshold value.

[0055] In the monitoring of the sudden leakage accident of marine chemicals, the present application can quickly determine whether the ultraviolet image of the chemicals floating on the water surface photographed by the ultraviolet camera is blurred, thereby avoiding manual screening of clear ultraviolet images by naked eyes and improving the efficiency of screening high-quality chemical ultraviolet images. Screening high-quality clear chemical ultraviolet images can provide high-quality image data for subsequent chemical target segmentation tasks and chemical target detection tasks, thereby improving the processing efficiency and accuracy of these tasks. In summary, the present application is of great significance for emergency monitoring of sudden leakage accidents of marine chemicals.

[0056] The above-described embodiments express the specific implementation of the present application, which is described in more detail and in more detail, and is intended to help understand the method of the present application and its core idea, but cannot be understood as a limitation on the scope of the patent of the present application. It should be noted that for ordinary skilled persons in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are all within the scope of protection of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A method for UV image blur detection based on DFT matrix integer approximate factorization, characterized in that, The method comprises the following steps: (1) The UV camera samples different shooting targets to obtain UV images corresponding to different targets; (2) A two-dimensional matrix corresponding to the UV image of the chemical is multiplied by a DFT matrix to perform discrete Fourier transform on the UV image to obtain a spectrum image; (3) The center of the spectrum image after the centering process and the low-frequency information of the image are shielded to complete high-pass filtering of the spectrum image; (4) The average high-frequency amplitude of the spectrum image after high-pass filtering is calculated to obtain the average high-frequency amplitude of the image as the blur value of the image for blur judgment; (5) The threshold of the image definition is set, and the blur value of the image is compared with the threshold. When the blur value of the image is less than the threshold, the image is determined to be a blurred image.

2. The UV image blur detection method based on DFT matrix integer-approximate factorization according to claim 1, characterized in that, The shooting target in the step (1) is a colorless chemical floating on the water surface, and the UV image spectrum range is 10-400 nm.

3. The UV image blur detection method based on DFT matrix integer-approximate factorization according to claim 1, characterized in that, In the step (2), the DFT matrix is obtained by integer approximate decomposition, and the matrix size is consistent with the size of the UV image.

4. The UV image blur detection method based on DFT matrix integer-approximate factorization according to claim 3, characterized in that, The integer approximate decomposition is specifically: the DFT matrix is decomposed into the form of multiplication of three sparse matrices, and the element values in the sparse matrices are all integers; the matrix representation is as follows: F N ≈A1A2A3; In the formula, F N is a DFT matrix, and A1, A2, and A3 are three sparse matrices obtained by integer approximate factorization.

5. The UV image blur detection method based on DFT matrix integer-approximate factorization according to claim 3, characterized in that, In the step (2), the discrete Fourier transform is performed on the UV image to obtain a spectrum image, which is specifically: the UV image of the chemical is multiplied by the three sparse matrices obtained after the integer approximate decomposition to perform the discrete Fourier transform to obtain the spectrum image of the UV image of the chemical; the matrix representation is as follows: X=A1A2A3x; In the formula, x is the matrix representation of the UV image of the chemical, and X is the spectrum image of the UV image after the discrete Fourier transform.

6. The method according to claim 1, wherein, In the step (3), the spectrum image after the centering process is specifically: the zero-frequency component of the spectrum image is moved to the center position of the spectrum image, and the low-frequency information is shielded by shielding the center region of the image based on the spectrum image.

7. The method according to claim 1, wherein, In the step (4), the average high-frequency amplitude of the image is obtained by statistically calculating the high-frequency amplitude of the whole image on the spectrum image after the low-frequency information is shielded.

8. The method according to claim 1, wherein, The threshold in the step (5) is manually set according to the quality of the UV image.

9. An ultraviolet image blur detection device based on DFT matrix integer approximation decomposition, characterized by, The device comprises the following modules: An image acquisition module: the UV camera samples different shooting targets to obtain UV images corresponding to different targets; A discrete Fourier transform module: a two-dimensional matrix corresponding to the UV image is multiplied by a DFT matrix to perform discrete Fourier transform on the UV image to obtain a spectrum image; A high-pass filtering module: the center of the spectrum image after the centering process and the low-frequency information of the image are shielded to complete high-pass filtering of the spectrum image; A blur value calculation module: the average high-frequency amplitude of the spectrum image after high-pass filtering is calculated to obtain the average high-frequency amplitude of the image as the blur value of the image for blur judgment; A blur detection module: the threshold of the image definition is set, and the blur value of the image is compared with the threshold. When the blur value of the image is less than the threshold, the image is determined to be a blurred image.

Images