A method for edge detection based on fourier single-pixel imaging

Abstract

The application discloses an edge detection method based on Fourier single-pixel imaging, comprising the following steps: S1, collecting zero-frequency optical signals; S2, calculating a modulus; S3, setting a high modulus coefficient threshold and an initial value of an error capacity; S4, obtaining a high-frequency information subset; S5, forming a Fourier spectrum; S6, performing spatial filtering on the Fourier spectrum; and S7, obtaining an edge detection result. The application discloses an edge detection method based on Fourier single-pixel imaging, realizes real-time optimization of a sampling process, improves a signal-to-noise ratio of edge detection, and solves the problem of fuzzy edge detection results. The application can be widely applied to biomedical imaging (such as quantum-classical fusion framework magnetic resonance imaging), industrial detection and security, remote control and surveying, visible light photography, three-dimensional imaging, hyperspectral imaging, and future automatic driving, emergency management, and ocean exploration and other frontier scientific research and industrial application fields.

Description

Technical Field

[0001] This invention relates to the fields of quantum computing and single-pixel imaging technology, specifically to an edge detection method based on Fourier single-pixel imaging. Background Technology

[0002] Quantum computing and quantum imaging technologies are key technologies for future industrial applications and also a major area for quantum technology applications that will benefit the general public. Single-pixel imaging is a novel computational imaging technology that utilizes spatial light modulation technology to obtain spatial information using only a single-pixel detector, without requiring spatial resolution.

[0003] The core of single-pixel imaging technology is to transform the spatial sampling process of the array detector into a set of spatial structure patterns and encode them into a programmable spatial light modulator. After the scattered light from the object is transmitted to the spatial light modulator, the spatial structure pattern continuously modulates the scattered light to obtain a modulation pattern. The modulation pattern is received by the single-pixel detector, which measures the intensity value of the object and reconstructs the target pattern based on the intensity value.

[0004] Single-pixel imaging is widely used in edge detection, such as in biomedical imaging (e.g., quantum-classical fusion framework magnetic resonance imaging), industrial inspection and security, remote control and surveying, visible light photography, 3D imaging, and hyperspectral imaging. In the future, it will be even more widely applied in cutting-edge scientific research such as autonomous driving, emergency management, and marine surveying. Compared to traditional edge detection, single-pixel imaging can directly inspect the edges of objects. In 2015, quantum imaging was first used to directly detect the edges of objects in the gradient domain; subsequently, the inspection technology of quantum imaging was further improved by using enhanced illumination modes, such as speckle shift mode, subpixel speckle mode, and interferometric structure mode. In addition, Fourier single-pixel imaging and Hadamard single-pixel imaging can also be used for edge detection.

[0005] However, current single-pixel imaging technology has poor image smoothing effect and insufficient noise removal, resulting in unclear actual edge features. Especially at low sampling rates, the image edge detection results are blurry and details are easily lost. Summary of the Invention

[0006] To address the problem of blurry edge detection results in current single-pixel imaging technology, this invention provides an edge detection method based on Fourier single-pixel imaging.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows: An edge detection method based on Fourier single-pixel imaging includes the following steps: S1: Acquire zero-frequency optical signals; S2: Calculate the modulus of the zero-frequency Fourier coefficients based on the zero-frequency optical signal; S3: Set the initial values ​​for the high-mode coefficient threshold and error capacity based on the magnitude of the zero-frequency Fourier coefficients; S4: Based on the high-mode coefficient threshold and error capacity, radial sampling is performed along each preset radial sampling path to obtain a high-frequency information subset of each radial sampling path; In any radial sampling path, Fourier coefficients are collected sequentially from zero frequency to high frequency region along the radial direction; If the magnitude F of the current Fourier coefficient is greater than the high-mode coefficient threshold T, then the current Fourier coefficient is determined to be a high-mode coefficient, and the error capacity is increased. If F≤T, then the current Fourier coefficients are classified as low-mode coefficients, and the error capacity is reduced. Radial sampling of the current radial sampling path is completed until the error capacity E ≤ T; S5: Form a Fourier spectrum based on the high-frequency information subsets of each radial sampling path; S6: Perform spatial filtering on the Fourier spectrum; S7: Obtain the edge detection result based on the filtered Fourier spectrum.

[0008] In the above scheme, by combining threshold judgment and error capacity control mechanism in the radial sampling path, the sampling process is optimized in real time, the signal-to-noise ratio of edge detection is improved, and the problem of unclear edge detection results is solved.

[0009] Preferably, in step S2, the magnitude of the zero-frequency Fourier coefficients is calculated using the following formula: ; in, , , and These are zero-frequency light signals collected under different phase illumination light fields.

[0010] Preferably, in step S3, the high-modulus coefficient threshold T = α * M is set; where α is a threshold adjustment parameter, and α > 0.

[0011] Preferably, in step S3, the initial value of the error capacity E is set to M.

[0012] Preferably, in step S4, if F > T, then the error capacity E = E + F is increased.

[0013] Preferably, in step S4, if F≤T, then the error capacity is reduced, E=E*β, where β is the error attenuation parameter, 0<β<1.

[0014] Preferably, in step S6, the Fourier spectrum is spatially filtered using a LoG filter.

[0015] Preferably, the LoG filter is represented as: ; in, For horizontal frequency components, For vertical frequency components, The standard deviation is denoted as .

[0016] Preferably, in step S7, the filtered Fourier spectrum is represented as follows: ; in, This is the Fourier spectrum.

[0017] The edge detection method based on Fourier single-pixel imaging according to claim 9 is characterized in that, in step S7, the edge detection result is obtained by performing an inverse Fourier transform on the filtered Fourier spectrum. ; in, This is the inverse Fourier transform operator.

[0018] Beneficial technical effects of the present invention: This invention provides an edge detection method based on Fourier single-pixel imaging. By combining threshold judgment and error capacity control mechanism in the radial sampling path, the sampling process is optimized in real time, improving the signal-to-noise ratio of edge detection and solving the problem of unclear edge detection results. Attached Figure Description

[0019] Fig. 1 This is a block diagram of the modules of the present invention; Fig. 2 This is a flowchart illustrating the implementation steps of the technical solution of the present invention; Fig. 3 This is a flowchart illustrating the steps of radial sampling in any radial sampling path in this invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the embodiments. However, the scope of protection of this invention is not limited to the specific embodiments described below. Example

[0021] like Figs. 1-3 As shown, an edge detection method based on Fourier single-pixel imaging includes the following steps: S1: Acquire zero-frequency optical signals; S2: Calculate the modulus of the zero-frequency Fourier coefficients based on the zero-frequency optical signal; S3: Set the initial values ​​for the high-mode coefficient threshold and error capacity based on the magnitude of the zero-frequency Fourier coefficients; S4: Based on the high-mode coefficient threshold and error capacity, radial sampling is performed along each preset radial sampling path to obtain a high-frequency information subset of each radial sampling path; In any radial sampling path, Fourier coefficients are collected sequentially from zero frequency to high frequency region along the radial direction; If the magnitude F of the current Fourier coefficient is greater than the high-mode coefficient threshold T, then the current Fourier coefficient is determined to be a high-mode coefficient, and the error capacity is increased. If F≤T, then the current Fourier coefficients are classified as low-mode coefficients, and the error capacity is reduced. Radial sampling of the current radial sampling path is completed until the error capacity E ≤ T; S5: Form a Fourier spectrum based on the high-frequency information subsets of each radial sampling path; S6: Perform spatial filtering on the Fourier spectrum; S7: Obtain the edge detection result based on the filtered Fourier spectrum.

[0022] In the specific implementation process, by combining threshold judgment and error capacity control mechanism in the radial sampling path, the sampling process was optimized in real time, the signal-to-noise ratio of edge detection was improved, and the problem of unclear edge detection results was solved. Example

[0023] An edge detection method based on Fourier single-pixel imaging includes the following steps: S1: Acquire zero-frequency optical signals; In practice, a spatial light modulator is used to generate a zero-frequency Fourier pattern, which is then used to illuminate the target object. The zero-frequency light signal is then collected by a single-pixel detector. S2: The processing module calculates the magnitude of the zero-frequency Fourier coefficients based on the zero-frequency optical signal. ; in, , , and These are zero-frequency light signals collected under different phase illumination light fields.

[0024] S3: The processing module sets the high-mode coefficient threshold T=α*M based on the modulus of the zero-frequency Fourier coefficients, and sets the initial value of the error capacity to M. Where α is the threshold adjustment parameter, α > 0; S4: The radial sampling module performs radial sampling along each preset radial sampling path based on the high-mode coefficient threshold and error capacity to obtain a subset of high-frequency information for each radial sampling path; In any radial sampling path, Fourier coefficients are collected sequentially from zero frequency to high frequency region along the radial direction; If F > T, then the current Fourier coefficients are determined to be high-mode coefficients, and the error capacity E = E + F is increased; If F≤T, then the current Fourier coefficients are determined to be low-mode coefficients, and the error capacity is reduced, E=E*β, where β is the error attenuation parameter, 0<β<1; Radial sampling of the current radial sampling path is completed until the error capacity E ≤ T; Typically, there are a few low-mode coefficients among the high-mode coefficients. By introducing an error capacity, sampling can continue after encountering low-mode coefficients. When high-mode coefficients are obtained, E increases, and when low-mode coefficients are obtained, E decreases. S5: Form a Fourier spectrum based on the high-frequency information subsets of each radial sampling path; In actual implementation, the processing module obtains the Fourier coefficients corresponding to each high-frequency point by performing a four-step phase shift method on the high-frequency information subset, and sums up the obtained Fourier coefficients to obtain the Fourier spectrum of the adaptive sampling distribution. S6: Spatial filtering of the Fourier spectrum is performed using a LoG filter; More specifically, the LoG filter is represented as: ; in, For horizontal frequency components, For vertical frequency components, The standard deviation is denoted as .

[0025] In the specific implementation process, specific information of the target object can be extracted by spatial filtering the Fourier spectrum of the target object; the LoG filter combines Gaussian smoothing and Laplacian edge detection. First, the image is Gaussian smoothed, and then the Laplacian operator is applied to the smoothed image, which can significantly improve the signal-to-noise ratio of edge detection at the same sampling rate.

[0026] S7: The processing module obtains the edge detection result based on the filtered Fourier spectrum.

[0027] More specifically, in step S7, the filtered Fourier spectrum is represented as follows: ; in, This is the Fourier spectrum.

[0028] More specifically, in step S7, the edge detection result is obtained by performing an inverse Fourier transform on the filtered Fourier spectrum: ; in, This is the inverse Fourier transform operator.

[0029] Based on the disclosure and teachings of the foregoing specification, those skilled in the art can make changes and modifications to the above embodiments. Therefore, the present invention is not limited to the specific embodiments disclosed and described above, and some modifications and changes to the invention should also fall within the protection scope of the claims of the present invention. Furthermore, although some specific terms are used in this specification, these terms are only for convenience of explanation and do not constitute any limitation on the present invention.

Claims

1. An edge detection method based on Fourier single-pixel imaging, characterized in that, Includes the following steps: S1: Acquire zero-frequency optical signals; S2: Calculate the modulus of the zero-frequency Fourier coefficients based on the zero-frequency optical signal; S3: Set the initial values ​​for the high-mode coefficient threshold and error capacity based on the magnitude of the zero-frequency Fourier coefficients; S4: Based on the high-mode coefficient threshold and error capacity, radial sampling is performed along each preset radial sampling path to obtain a high-frequency information subset of each radial sampling path; In any radial sampling path, Fourier coefficients are collected sequentially from zero frequency to high frequency region along the radial direction; If the magnitude F of the current Fourier coefficient is greater than the high-mode coefficient threshold T, then the current Fourier coefficient is determined to be a high-mode coefficient, and the error capacity is increased. If F≤T, then the current Fourier coefficients are classified as low-mode coefficients, and the error capacity is reduced. Radial sampling of the current radial sampling path is completed until the error capacity E ≤ T; S5: Form a Fourier spectrum based on the high-frequency information subsets of each radial sampling path; S6: Perform spatial filtering on the Fourier spectrum; S7: Obtain the edge detection result based on the filtered Fourier spectrum.

2. The edge detection method based on Fourier single-pixel imaging according to claim 1, characterized in that, In step S2, the magnitude of the zero-frequency Fourier coefficients is calculated using the following formula: ； in, , , and These are zero-frequency light signals collected under different phase illumination light fields.

3. The edge detection method based on Fourier single-pixel imaging according to claim 2, characterized in that, In step S3, the high-modulus coefficient threshold T = α * M is set; where α is the threshold adjustment parameter, and α > 0.

4. The edge detection method based on Fourier single-pixel imaging according to claim 2, characterized in that, In step S3, the initial value of the error capacity E is set to M.

5. The edge detection method based on Fourier single-pixel imaging according to claim 1, characterized in that, In step S4, if F > T, then increase the error capacity E = E + F.

6. The edge detection method based on Fourier single-pixel imaging according to claim 1, characterized in that, In step S4, if F≤T, then the error capacity is reduced, E=E*β, where β is the error attenuation parameter, 0<β<1.

7. The edge detection method based on Fourier single-pixel imaging according to claim 1, characterized in that, In step S6, the Fourier spectrum is spatially filtered using a LoG filter.

8. The edge detection method based on Fourier single-pixel imaging according to claim 7, characterized in that, The LoG filter is represented as follows: ； in, For horizontal frequency components, For vertical frequency components, The standard deviation is denoted as .

9. The edge detection method based on Fourier single-pixel imaging according to claim 8, characterized in that, In step S7, the filtered Fourier spectrum is represented as follows: ； in, This is the Fourier spectrum.

10. The edge detection method based on Fourier single-pixel imaging according to claim 9, characterized in that, In step S7, the edge detection result is obtained by performing an inverse Fourier transform on the filtered Fourier spectrum. ； in, This is the inverse Fourier transform operator.

Images