Turbid underwater object imaging calculation method based on logistic regression and storage medium

Abstract

This invention discloses a logistic regression-based method for imaging objects in turbid water and a storage medium thereof. The method includes the following steps: acquiring image data using the Hadamard single-pixel imaging method; processing the acquired data using the Hadamard single-pixel imaging method in two steps: first, multinomial fitting; second, spectral data feature enhancement; and finally, performing an inverse Hadamard transform on the data obtained after the above two processing steps to obtain the final image. This method reduces the computational cost of traditional turbid water imaging by modifying only one data point in the spectrum, thus significantly improving the contrast of the final image. This greatly reduces the computational burden of image enhancement, resulting in a three- to five-fold increase in contrast compared to the original image. Furthermore, it maintains good performance even in highly turbid water and provides good imaging results for objects made of materials such as paper, plastic, and metal, demonstrating broad adaptability and high robustness.

Description

Technical Field

[0001] This invention relates to the field of underwater imaging technology, specifically to a computational method for imaging objects in turbid water based on logistic regression. Background Technology

[0002] Traditional optical technologies, such as CCDs and CMOS sensors, suffer from blurring and reduced contrast when imaging turbid underwater targets. This is due to light attenuation caused by water absorption and scattering caused by the inhomogeneity of the medium in turbid water. Existing methods for improving the imaging quality of objects in turbid water mainly fall into two categories: image restoration methods and image enhancement methods. Image restoration methods often require extremely complex physical models and expensive imaging structures, and are mostly only suitable for one or a few specific environments. Image enhancement methods, on the other hand, require a huge amount of computation. Although both methods can improve the quality of images of objects in turbid water to some extent, they still lag far behind images in clear water and have various limitations, especially in highly turbid water, where the results are often unsatisfactory.

[0003] Currently, there are two main categories of methods for improving the imaging quality of objects underwater in turbid water: image restoration methods and image enhancement methods. Image restoration methods often require the establishment of extremely complex physical models and expensive imaging structures, and most of them can only be adapted to one or a few specific environments. Image enhancement methods require a huge amount of computation and are not well adapted to different materials, water qualities, and turbidity levels. Especially in highly turbid water, the current methods are not satisfactory. Summary of the Invention

[0004] The present invention proposes an imaging calculation method for turbid underwater objects based on logistic regression, which can at least solve one of the above-mentioned technical problems.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] An image processing method for underwater objects in turbid water based on logistic regression includes the following steps:

[0007] Image data acquisition was performed using the Hadama single-pixel imaging method.

[0008] The data acquired using the Hadamard single-pixel imaging method is processed in two steps: the first step is polynomial fitting, and the second step is spectral data feature enhancement.

[0009] Finally, the data obtained after the above two processing steps are subjected to inverse Hadamard transform to obtain the final image.

[0010] Furthermore, the image data acquisition using the Hadamard single-pixel imaging method specifically includes,

[0011] set up:

[0012] P H (x, y) = {δ H (u, v)} (1)

[0013]

[0014] Where (x, y) are the coordinates in the image space domain, (u, v) are the coordinates in the Hadamard domain, and the operator H... -1 Represents the inverse Hadamard transformation;

[0015] Since there are no negative values ​​in the DMD or the image, the Hadamard single-pixel imaging method requires differential operations; the corresponding Hadamard spectrum H s The formula is as follows:

[0016] H s (u, v) = D + -D - (3)

[0017] Where D + and D _ These correspond to the object at P respectively H+ and P H- Measurements under illumination.

[0018] Furthermore, the polynomial fitting steps are as follows:

[0019] First, the collected data is expanded to powers from 0 to 4 to prepare for the next step of processing, as shown in the following formula:

[0020]

[0021] Where x represents the collected data, and t represents the turbidity of the underwater environment at the time the image data was collected; then, the integrated data is subjected to the following polynomial fitting process:

[0022]

[0023] Where Y represents the output, W represents the weight matrix, and the fitting target value is the data R of the corresponding image under clear water.

[0024] W * =argmin||W||2,subject to XW=R (6)

[0025] Furthermore, the spectrum data feature enhancement processing includes,

[0026] The spectral information corresponding to each image is processed using the following formula:

[0027]

[0028] That is, only the point (1,1) in the frequency domain is processed, and the normalized Hadamard matrix is ​​mapped to the entire image during the subsequent inverse transformation.

[0029] On the other hand, the present invention also discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the method described above.

[0030] As can be seen from the above technical solutions, the core of the logistic regression-based underwater object imaging calculation method of the present invention lies in (1) using the differential Hadamard single-pixel imaging method to collect one-dimensional data, which to a certain extent suppresses the backscattering effect of turbid water; (2) using a fourth-order polynomial to fit the one-dimensional data collected by the Hadamard single-pixel imaging method, so that the object imaging data under turbid water is close to the data in clear water; (3) using formula 11 to enhance the data features in the spectral domain of the data obtained after fitting, which improves the contrast of the image and improves the problem of unclear overall image information and low contrast caused by the absorption of light by the water. Moreover, compared with other methods that use single-pixel imaging methods for imaging and then perform post-processing, the method of the present invention only needs to process one point in the spectral domain before imaging, which greatly reduces the amount of data processed.

[0031] The logistic regression-based underwater object imaging calculation method of this invention reduces the cost of traditional underwater imaging in turbid water. It only modifies a single data point on the spectrum, thereby significantly improving the contrast of the final image and greatly reducing the computational load of image enhancement. The contrast of the image is improved by three to five times compared to the original image. In addition, it can still achieve good results in highly turbid water (greater than 50 NTU) and has good imaging effects on objects made of paper, plastic, metal and other materials. It has wide adaptability and high robustness. Attached Figure Description

[0032] Figure 1 This is a network architecture diagram of the method of the present invention;

[0033] Figure 2 This is a comparison chart of the results of an embodiment of the present invention. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0035] like Figure 1 The diagram shown is the network framework of the method. The image data acquisition method uses Hadamard single-pixel imaging, and the principle of Hadamard single-pixel imaging is as follows:

[0036] P H (x, y) = H -1 {δ H (u,u)} (1)

[0037]

[0038] Where (x, y) are the coordinates in the image space domain, and (u, v) are the coordinates in the Hadamard domain. Operator H -1 This represents the inverse Hadamard transform. Since there are no negative values ​​in the DMD or image, the Hadamard single-pixel imaging method requires difference operations. The corresponding Hadamard spectrum H... s The formula is as follows:

[0039] H s (u, v) = D + -D - (3)

[0040] Where D + and D _ These correspond to the object at P respectively H+ and P H- Measurements under illumination. Because the data was acquired using a differential Hadamard single-pixel imaging method, it has a certain effect on suppressing backscattering in underwater turbid media.

[0041] Subsequently, the data acquired using the Hadamard single-pixel imaging method is processed in two steps: the first step is polynomial fitting, and the second step is spectral data feature enhancement.

[0042] In the first step of polynomial fitting, this invention first expands the collected data to powers from 0 to 4 to prepare for the next step of processing, as shown in the following formula:

[0043]

[0044] Where x represents the collected data, and t represents the turbidity of the underwater environment at the time the image data was collected. Then, the integrated data is subjected to the following polynomial fitting process;

[0045]

[0046] Where Y represents the output and W represents the weight matrix. The fitting target value is the data R of the corresponding image under clear water.

[0047] W * =argmin||W||2,subject to XW=R (6)

[0048] After this step, the problems of forward scattering, backscattering, and light absorption by water are comprehensively improved, enhancing the overall image quality. Following this step, a second step of spectral data feature enhancement is performed. This step is possible only if the normalized Hadamard matrix was used to generate the projection image when acquiring data using the Hadamard single-pixel imaging method. Under these conditions, the present invention uses the following formula to process the spectral information corresponding to each image as follows:

[0049]

[0050] In other words, by processing only the point (1,1) in the spectral domain, the normalized Hadamard matrix can be mapped to the entire image during the subsequent inverse transform. This step greatly reduces the problem of the overall dark image and low contrast caused by water absorption, thereby improving the image contrast.

[0051] Finally, the data obtained after the above two processing steps is subjected to inverse Hadamard transform to obtain the final image. The imaging result is as follows. Figure 2 As shown;

[0052] Figure 2 Image (a) shows the results obtained directly using a traditional CMOS camera under three different turbidity levels. Image (b) shows the result of (a) after histogram equalization. Image (c) shows the result of (a) after image grayscale adjustment. Image (d) shows the result of direct Hadamard single-pixel imaging. Image (e) shows the result after processing using the method proposed in this invention. The results show that the image contrast is greatly improved after using the method of this invention. The imaging effect is significantly better than the results obtained by traditional CMOS camera shooting and its processed form. Furthermore, the results show that the method of this invention still has good performance in highly turbid water environments (above 50 NTU).

[0053] In summary, the core of this invention lies in (1) using the differential Hadamard single-pixel imaging method to acquire one-dimensional data, which to a certain extent suppresses the backscattering effect of turbid water; (2) using a fourth-order polynomial to fit the one-dimensional data acquired by the Hadamard single-pixel imaging method, making the image data of objects in turbid water close to that in clear water; (3) using Formula 11 to enhance the data features in the spectral domain of the data obtained after fitting, thereby improving the contrast of the image and addressing the problem of unclear overall image information and low contrast caused by the absorption of light by the water. Moreover, compared with other methods that use single-pixel imaging methods for imaging and then perform post-processing, the method of this invention only needs to process one point in the spectral domain before imaging, greatly reducing the amount of data processed.

[0054] The key point of this invention is that when using the Hadamard single-pixel imaging method to acquire data, the acquired data is fitted with a fourth-order polynomial and the spectral domain data feature is enhanced using Formula 11 to improve image contrast and improve the quality of underwater object imaging in turbid water.

[0055] In summary, existing methods for imaging objects in turbid water often require extremely complex models and expensive imaging equipment or involve enormous computational loads. Furthermore, they are often only applicable to one or a few specific environments or objects made of specific materials, and their imaging results are often unsatisfactory in highly turbid water environments. In contrast, the method of this invention has a simple structure, uses inexpensive imaging equipment, requires very little data processing after learning the corresponding weight values, and can adapt to various imaging environments and objects made of different materials (such as paper, plastic, and iron). More importantly, the method of this invention also achieves excellent imaging results for objects in highly turbid water environments.

[0056] In another aspect, the present invention also discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of any of the methods described above.

[0057] In another aspect, the present invention also discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of any of the methods described above.

[0058] In another embodiment provided in this application, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to perform the steps of any of the methods described in the above embodiments.

[0059] It is understood that the system provided in the embodiments of the present invention corresponds to the method provided in the embodiments of the present invention, and the explanation, examples and beneficial effects of the relevant content can be referred to the corresponding parts of the above method.

[0060] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0061] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0062] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A computational method for imaging objects in turbid water based on logistic regression, characterized in that, Includes the following steps, Image data acquisition was performed using the Hadama single-pixel imaging method. The data acquired using the Hadamard single-pixel imaging method is processed in two steps: the first step is polynomial fitting, and the second step is spectral data feature enhancement. Finally, the data obtained after the above two processing steps are subjected to inverse Hadamard transform to obtain the final image. The image data acquisition using the Hadamard single-pixel imaging method specifically includes, set up: in These are the coordinates in the image spatial domain. The coordinates of the Hadamah domain, operator Represents the inverse Hadamard transformation; Since there are no negative values ​​in the DMD or the image, the Hadamard single-pixel imaging method requires differential operations; the corresponding Hadamard spectrum The formula is as follows: in and Corresponding to the object in and Measurements under illumination; The polynomial fitting steps are as follows: First, the collected data is expanded to powers from 0 to 4 to prepare for the next step of processing, as shown in the following formula: in For the collected data, The underwater environment turbidity at the time the image data was acquired was used as the basis for the following polynomial fitting process: in Indicates the output. The weight matrix represents the weight values; the target value for fitting is R, which represents the data of the corresponding image in clear water. ； The spectrum data feature enhancement processing includes, The spectral information corresponding to each image is processed using the following formula: That is, only for the spectral domain This single point is processed, and the normalized Hadamard matrix is ​​mapped to the entire image during the subsequent inverse transform.

2. A computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to perform the steps of the method as claimed in claim 1.

Images