Weight function and multi-exposure fusion-based analysis and detection method and device

Abstract

The invention discloses an analysis and detection method and device based on a weighting function and multi-exposure fusion, and belongs to the field of immunochromatography analysis and detection. The analysis and detection method comprises the following steps: obtaining a multi-exposure image set under a plurality of multi-exposure time sequences, in the process of recovering a camera response function of an image acquisition system, using an HDR algorithm based on multi-exposure fusion, adjusting weights of different exposure images by adopting a segmented weight function, and accurately adjusting and controlling parameters of the segmented weight function. The characteristics of each concentration section in immunochromatography are reflected more finely, and an optimal weighting function is established to enhance the fitting accuracy of a camera response function, so that an immunochromatography HDR image with the highest quality is recovered, and the linear range of quantitative detection is expanded.

Description

Technical Field

[0001] The present invention belongs to the field of immunochromatographic analysis and detection, and more specifically, relates to an analysis and detection method and device based on weight function and multi-exposure fusion. Background Art

[0002] As an easy-to-operate on-site rapid detection technology, the immunochromatography (lateral flow immunoassay LFIA) method has been widely used in many fields. Rapid test reagent cards represented by colloidal gold immunochromatography cards can directly interpret positive and negative results with the naked eye.

[0003] In recent years, with the rapid development of mobile devices such as handheld devices and smart phones, using the cameras of these portable devices for immunochromatographic image detection has become a new trend. These immunoassay devices based on the principle of image imaging can quickly, conveniently and cost-effectively capture all image information of the entire test strip card, providing a basis for subsequent data processing and global analysis.

[0004] Due to the limited dynamic range of images captured by the camera, it is easy to have problems with bright areas being overexposed and whitened or dark areas losing details when taking images of immunochromatographic test strips, which seriously affects the accuracy and reproducibility of the test results. For fluorescent immunochromatographic technology, which requires higher sensitivity and a wider signal range, the defect of the image imaging detection equipment having a narrow dynamic range is more prominent. In order to take into account the dynamic range of detection, many devices have to sacrifice the advantages of size and cost and use a more cumbersome and expensive mechanical scanning chromatography detection method.

[0005] In the field of consumer digital cameras, the development of HDR technology has expanded the dynamic range of photos, solved the problem of loss of dark or brightness details in photos, and allowed a photo to show both bright and dark parts. However, traditional HDR technology is committed to showing the details of highlight areas and shadow areas in a photo at the same time, and mainly focuses on improving the contrast and visual effects of the image. It does not pay attention to the precise correspondence between the synthesis result and the actual brightness value, nor does it consider the signal distortion caused by nonlinear conversion. In immunochromatography image detection, it is necessary not only to retain the light and dark details in the image, but more importantly to ensure that the image can accurately reflect the true reaction intensity of each area on the test paper. In addition, traditional HDR technology usually keeps the light source conditions fixed and takes a set of photos with different exposure times in the same light source scene.

[0006] Due to the limited dynamic range of photos in traditional HDR technology, the immunochromatographic detector based on the image imaging principle has problems such as narrow detection range, slow detection speed and low detection accuracy. Summary of the invention

[0007] In view of the defects of the related art, the purpose of the present invention is to provide an analysis and detection method based on weight function and multi-exposure fusion, aiming to solve the problem that the dynamic range of photos in traditional HDR technology is limited, which leads to the problem that the immunochromatographic detector based on the image imaging principle has a narrow detection range, slow detection speed and low detection accuracy.

[0008] To achieve the above objectives, in a first aspect, the present invention provides an analysis and detection method based on a weight function and multi-exposure fusion, comprising:

[0009] Shooting the paper card to be tested in a multi-exposure time sequence to obtain a multi-exposure image set of the paper card to be tested;

[0010] According to the standard concentration quantitative curve, the multi-exposure image set of the paper card to be tested and the corresponding exposure time, the actual concentration value of the paper card to be tested is obtained;

[0011] The method for generating the standard concentration quantitative curve comprises:

[0012] S1, photographing P multi-exposure time sequences of X standard immunochromatographic test paper cards of different concentrations to obtain a multi-exposure image set; dividing the multi-exposure image set into a first multi-exposure image set including X1 standard immunochromatographic test paper card images and a second multi-exposure image set including X2 standard immunochromatographic test paper card images; wherein P≥3, X=X1+X2;

[0013] S2. Obtain grayscale value changes of N pixels in the T-line area of ​​each image in the first multi-exposure image set to obtain a grayscale value matrix of N×P pixels;

[0014] S3. Construct the following piecewise weight function:

[0015]

[0016] Wherein, the function forms of f1(z) and f2(z) include linear function, quadratic polynomial function, exponential function or logarithmic function, and z1 and z2 are segmentation points;

[0017] S4, determining the function forms of f1(z) and f2(z) and the segmentation points z1 and z2 that minimize the error between the average pixel value of the T-line area and the true pixel value, and obtaining a target segmentation weight function; the calculation method of the average pixel value of the T-line area includes: obtaining a camera response function according to the segmentation weight function combined with a gray value matrix; obtaining an HDR image according to the camera response function, a second multi-exposure image set and a corresponding exposure time; performing tone mapping on the HDR image to obtain an average pixel value of the T-line area;

[0018] S5, obtaining a target camera response function according to the target segment weight function combined with the gray value matrix;

[0019] S6. According to the target camera response function, the multi-exposure image set of the standard immunochromatographic test paper card and the corresponding exposure time, the average pixel value of the T-line area of ​​the standard immunochromatographic test paper card is obtained, and the average pixel value of the T-line area of ​​the standard immunochromatographic test paper card is fitted with the corresponding concentration to generate a standard concentration quantitative curve.

[0020] Optionally, the standard immunochromatographic test paper cards have different T line concentrations and the same C line concentrations.

[0021] Optionally, S2 specifically includes:

[0022] S21, extracting a T-line region of each image in the first multi-exposure image set using an image segmentation algorithm;

[0023] S22, vertically stitching the T-line regions of each image in order of concentration to obtain a guide image;

[0024] S23, acquiring grayscale value changes of N / X1 pixels in each T-line region in the guide image to obtain a grayscale value matrix of N×P pixels.

[0025] Optionally, the error between the average pixel value and the true pixel value is the root mean square error RMSE and the mean absolute percentage error MAPE of the average pixel value of the T-line area and the true pixel value of the standard immunochromatographic test paper card.

[0026] Optionally, S6 specifically includes:

[0027] S61, obtaining an average pixel value of a T-line region of the standard immunochromatographic test paper card according to a target camera response function, a multi-exposure image set of a standard immunochromatographic test paper card, and a corresponding exposure time;

[0028] S62. With the average pixel value in S61 as the y-axis and the actual concentration of the standard immunochromatographic test paper card as the x-axis, a four-parameter logistic fit is used to obtain a standard concentration quantitative curve.

[0029] Optionally, the standard immunochromatographic test paper card includes a colloidal gold immunochromatographic card and a fluorescent immunochromatographic card.

[0030] Optionally, the multi-exposure image set covers underexposure to overexposure in different exposure time sequences.

[0031] Optionally, after obtaining the multi-exposure image set, the method further includes: preprocessing the multi-exposure image set using a median filtering algorithm, a Gaussian filtering algorithm, a bilateral filtering algorithm or a guided filtering algorithm.

[0032] In a second aspect, the present invention further provides an analysis and detection device based on a weight function and multi-exposure fusion, which is used to perform the analysis and detection method as described in any one of the first aspects, including:

[0033] An image acquisition module is used to photograph the paper card to be tested in a multi-exposure time sequence to obtain a multi-exposure image set of the paper card to be tested;

[0034] The concentration analysis and detection module is used to obtain the actual concentration value of the paper card to be tested according to the standard concentration quantitative curve, the multi-exposure image set of the paper card to be tested and the corresponding exposure time;

[0035] Wherein, the method for generating the standard concentration quantitative curve includes:

[0036] Shooting P multi-exposure time sequences of X standard immunochromatographic test paper cards of different concentrations to obtain a multi-exposure image set; dividing the multi-exposure image set into a first multi-exposure image set including X1 standard immunochromatographic test paper card images and a second multi-exposure image set including X2 standard immunochromatographic test paper card images; wherein P≥3, X=X1+X2;

[0037] Acquire gray value changes of N pixels in the T-line region of each image in the first multi-exposure image set to obtain a gray value matrix of N×P pixels;

[0038] Construct the following piecewise weight function:

[0039]

[0040] Wherein, the function forms of f1(z) and f2(z) include linear function, quadratic polynomial function, exponential function or logarithmic function, and z1 and z2 are segmentation points;

[0041] Determine the function form of f1(z) and f2(z) and the segmentation points z1 and z2 that minimize the error between the average pixel value of the T-line area and the true pixel value, and obtain the target segmentation weight function; the calculation method of the average pixel value of the T-line area includes: obtaining a camera response function according to the segmentation weight function combined with the gray value matrix; obtaining an HDR image according to the camera response function, the second multi-exposure image set and the corresponding exposure time; and performing tone mapping on the HDR image to obtain the average pixel value of the T-line area;

[0042] The target camera response function is obtained according to the target segment weight function combined with the gray value matrix;

[0043] According to the target camera response function, the multi-exposure image set of the standard immunochromatographic test strip card and the corresponding exposure time, the average pixel value of the T-line area of ​​the standard immunochromatographic test strip card is obtained, and the average pixel value of the T-line area of ​​the standard immunochromatographic test strip card is fitted with the corresponding concentration to generate a standard concentration quantitative curve.

[0044] Compared with the prior art, the above technical solution conceived by the present invention can achieve the following beneficial effects:

[0045] 1. The present invention provides an analysis and detection method based on weight function and multi-exposure fusion. Aiming at the specific needs of quantitative processing of immunochromatographic images, based on the idea of ​​multi-exposure fusion high dynamic range imaging technology, the camera response function of the image acquisition system is first restored, and then the camera response function curve is combined with the biological response curve, thereby expanding the linear range of the immunochromatographic standard quantitative curve to solve the current limitation of the narrow detection range of the immunochromatographic detector based on the principle of image imaging. In the process of restoring the camera response function of the image acquisition system, the weights of images with different exposures are adjusted by using a segmented weight function to reduce the influence of over-exposed and under-exposed areas in the image on the fitting accuracy of the camera response function, so that the standard quantitative curve obtained by fitting fits the characteristics of the immunochromatographic S-curve, which can significantly improve the accuracy of HDR image synthesis and make the analysis and detection results based on the immunochromatographic test strip card more accurate.

[0046] 2. The present invention provides an analysis and detection method based on a weight function and multi-exposure fusion. After generating a standard concentration quantitative curve, a multi-exposure image set of a paper card to be tested is captured in a multi-exposure time sequence, thereby obtaining a high-precision HDR image and an accurate average pixel value of the T-line area of ​​the paper card to be tested. The actual concentration value of the paper card to be tested can be obtained by referring to the standard concentration quantitative curve. This method can be applied to colloidal gold immunochromatographic test paper cards and fluorescent immunochromatographic test paper cards, and accurate concentration detection results can be obtained. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a flowchart of an analysis and detection method based on a weight function and multi-exposure fusion provided by an embodiment of the present invention;

[0048] Figure 2 This is an example of a standard immunochromatographic test strip card;

[0049] Figure 3 It is 36 standard colloidal gold immunochromatographic test strip cards;

[0050] Figure 4 A collection of multi-exposure photos of different immunochromatographic test strips at 9 exposure times;

[0051] Figure 5(A) shows the images of 8 immunochromatographic test strips. Figure 5 Middle (B) is a guide image formed by combining the T-line area of ​​8 immunochromatographic test strip images. Figure 5 Middle (C) is to select Q pixels in each ROI area;

[0052] Figure 6 The results of the 5# test paper card under four filtering algorithms are shown in Figure 2, where (A) is the original image of the 5# test paper card, (B) is the bilateral filtering result, (C) is the median filtering result, (D) is the Gaussian filtering result, and (E) is the guided filtering result.

[0053] Figure 7 It is a flow chart of the sliding window threshold segmentation algorithm;

[0054] Figure 8 (A) is the sliding window threshold segmentation algorithm to extract the C line and T line of the immunochromatographic test paper card. Figure 8 (B) is the area with the minimum gray value selected by the sliding window of the ROI image of the immunochromatography image. Figure 8 Middle (C) shows the C-line and T-line regions after the sliding window segmentation;

[0055] Fig. 9 (A) is the exposure offset under the identity function. Fig. 9 (B) is the exposure offset under the hat function. Fig. 9 Middle (C) is the exposure offset under the Gaussian function;

[0056] Fig.10 The camera response function results fitted under different weight functions; where (A) is the identity function, (B) is the hat function, (C) is the Gaussian function, parameters a=2, b=4, (D) is the CRF curve fitted under the identity function, (E) is the CRF curve fitted under the hat function, and (F) is the CRF curve fitted under the Gaussian function;

[0057] Fig.11 (A) is the MAPE of different (a, b) parameter combinations. Fig.11 (B) RMSE of different (a, b) parameter combinations;

[0058] Fig.12 is the image of the Gaussian function when the parameters a=10, b=4;

[0059] Fig.13 (A) is the quantitative curve of 36 test strips divided into 9 groups and spliced ​​at 9 exposure times. Fig.13 Middle (B) is the slope of the mapping between the T-line pixel value and the actual printed grayscale value of the 9 groups of test strips;

[0060] Fig.14(A) The optimal piecewise function form, Fig.14 Middle (B) is the camera response function curve based on the optimal weight function fitting;

[0061] Fig.15 It is the standard concentration quantitative curve fitted under the HDR algorithm based on the target weight function;

[0062] Fig.16 This is a comparison of the measurable signal area of ​​the immunochromatographic test strip before and after the multi-exposure fusion HDR algorithm;

[0063] Fig.17 Comparison of the quantitative curves of standard concentrations before and after the handheld detector uses the HDR algorithm. DETAILED DESCRIPTION

[0064] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0065] The contents involved in the above embodiment are described below in conjunction with a preferred embodiment.

[0066] The immunochromatographic image sensor in the image acquisition system is essentially a special camera, and its basic working principle is the same as that of the digital cameras used in daily life. Therefore, the photo enhancement technology of digital cameras can be used to solve the dynamic range expansion problem of the immunochromatographic detection system. During immunochromatographic detection, it is necessary to flexibly adjust different parameters in the experiment according to the needs of the detection object, such as light source brightness, exposure time, etc., so as to produce a series of images under different brightness conditions, which helps to more effectively restore the most realistic immunochromatographic HDR images in immunochromatographic detection, surpassing the images obtained under traditional HDR technology.

[0067] An analysis and detection method based on weight function and multi-exposure fusion, comprising:

[0068] Shooting the paper card to be tested in a multi-exposure time sequence to obtain a multi-exposure image set of the paper card to be tested;

[0069] According to the standard concentration quantitative curve, the multi-exposure image set of the paper card to be tested and the corresponding exposure time, the actual concentration value of the paper card to be tested is obtained;

[0070] The method for generating the standard concentration quantitative curve comprises:

[0071] S1, photographing P multi-exposure time sequences of X standard immunochromatographic test paper cards of different concentrations to obtain a multi-exposure image set; dividing the multi-exposure image set into a first multi-exposure image set including X1 standard immunochromatographic test paper card images and a second multi-exposure image set including X2 standard immunochromatographic test paper card images; wherein P≥3, X=X1+X2;

[0072] S2. Obtain grayscale value changes of N pixels in the T-line area of ​​each image in the first multi-exposure image set to obtain a grayscale value matrix of N×P pixels;

[0073] S3. Construct the following piecewise weight function:

[0074]

[0075] Wherein, the function forms of f1(z) and f2(z) include linear function, quadratic polynomial function, exponential function or logarithmic function, and z1 and z2 are segmentation points;

[0076] S4, determining the function forms of f1(z) and f2(z) and the segmentation points z1 and z2 that minimize the error between the average pixel value of the T-line area and the true pixel value, and obtaining a target segmentation weight function; the calculation method of the average pixel value of the T-line area includes: obtaining a camera response function according to the segmentation weight function combined with a gray value matrix; obtaining an HDR image according to the camera response function, a second multi-exposure image set and a corresponding exposure time; performing tone mapping on the HDR image to obtain an average pixel value of the T-line area;

[0077] S5, obtaining a target camera response function according to the target segment weight function combined with the gray value matrix;

[0078] S6. According to the target camera response function, the multi-exposure image set of the standard immunochromatographic test paper card and the corresponding exposure time, the average pixel value of the T-line area of ​​the standard immunochromatographic test paper card is obtained, and the average pixel value of the T-line area of ​​the standard immunochromatographic test paper card is fitted with the corresponding concentration to generate a standard concentration quantitative curve.

[0079] like Figure 1 As shown in the figure, the specific process can be divided into (1) initialization stage: establishing and optimizing the camera response function (CRF) curve of the image acquisition device; (2) use stage: using the image acquisition device to capture a multi-exposure image set of a standard immunochromatographic test strip card with different concentrations, using the established camera response function curve to guide the multi-exposure image set to generate an HDR image, and then extracting the characteristic value of the region of interest (ROI) in the HDR image to expand the linear range of the standard quantitative curve.

[0080] Before acquiring the multi-exposure image set, also include:

[0081] Design X standard immunochromatographic test strips with different T line concentrations, and set the C line concentration of each standard immunochromatographic test strip to be the same. The standard immunochromatographic test strips include colloidal gold immunochromatographic cards and fluorescent immunochromatographic cards.

[0082] When the standard immunochromatographic test paper card is a colloidal gold immunochromatographic card, image processing software is used to create Figure 2 The standard immunochromatographic test strip card shown has a layout designed with a quality control strip area frame of corresponding size, including the quality control line (C line) and the test line (T line). The size of the quality control strip and the positions of the C line and T line are cut to be exactly the same as the size and position of the actual test strip card product, and three test strips of the same concentration are made at a time, reserved for subsequent repetitive tests.

[0083] After the standard colloidal gold immunochromatographic test paper card is designed, it is printed on Pantech 230g enhanced rough paper by a high-resolution color photocopying device. In this embodiment, a standard colloidal gold immunochromatographic test paper card with 36 different T-line grayscales with a pixel difference of 7 is designed, and the pixel value of the C line of each test paper card is the same. Figure 3 As shown, it represents 36 standard colloidal gold immunochromatographic test paper cards with the same C line gray value and different T line gray values.

[0084] When the standard immunochromatographic test paper card is a fluorescent immunochromatographic test paper card, the standard fluorescent immunochromatographic test paper card is composed of 19 self-made fluorescent immunochromatographic test paper cards with different T line concentrations. The T line concentrations of the 19 fluorescent immunochromatographic test paper cards range from low to high, covering from 101-10 4 The fluorescence intensity is high and the dynamic range is wide.

[0085] This embodiment uses a standard colloidal gold immunochromatographic test paper card as an example. The measurement of the standard colloidal gold immunochromatographic test paper card is carried out under dark conditions. A sample set of 36 test paper cards is photographed under a specific exposure to cover various conditions from underexposure to overexposure. The multi-exposure image set covers 0-255 pixel values ​​under different exposure time sequences. In this embodiment, P is set to 9, such as Figure 4As shown, the standard grayscale test paper card was shot at 9 exposure times (0.122ms, 0.244ms, 0.488ms, 0.977ms, 1.953ms, 3.906ms, 7.812ms, 15.625ms, 31.25ms), and a set of 36×9 multi-exposure photos from underexposure to overexposure was obtained. Eight of the test paper cards were evenly selected to cover the grayscale value range of 0-255 as much as possible, namely 1# test paper card, 5# test paper card, 10# test paper card, 15# test paper card, 20# test paper card, 25# test paper card, 30# test paper card, and 35# test paper card. Figure 5 As shown in (A), these 8×9 images are the first multi-exposure image set, which will be used to establish the camera response function curve.

[0086] After obtaining the multi-exposure image set, the image needs to be preprocessed. The acquired immunochromatographic images are collected by CMOS image sensors, and these images often contain noise, and the shorter the exposure time, the more noise points there are in the image. Figure 6 As shown, the median filter algorithm, Gaussian filter algorithm, bilateral filter algorithm or guided filter algorithm can be used to preprocess the multi-exposure image set. In this embodiment, the most effective guided filter algorithm is used to preprocess the exposure image set, which can effectively reduce the noise in the image and enhance the visibility of the target.

[0087] Optionally, S2 specifically includes:

[0088] S21 . Extracting a T-line region of each image in the first multi-exposure image set using an image segmentation algorithm.

[0089] The HDR image (first multi-exposure image set) in .hdr format is converted into a high-quality LDR image through a tone mapping algorithm to adapt to the dynamic range of conventional display devices. Determine the position and size of the region of interest (such as the T-line area) in the standard colloidal gold immunochromatographic test strip card, extract the ROI area, and convert it into a single-channel grayscale image. Divide the image into two parts, upper and lower, at the vertical midpoint of the grayscale image, and apply sliding window threshold segmentation to the two parts. The window moves along the vertical direction of the image, and calculates the average pixel value in each window. By comparing the pixel values ​​of different windows, find the area with the smallest pixel value, which corresponds to the T line and C line. The process of the sliding window threshold segmentation algorithm is as follows Figure 7 shown.

[0090] The segmented image is Figure 8 As shown, the T-line and C-line areas are clearly identified, and the average pixel value of the two areas is calculated.

[0091] As an alternative, other threshold segmentation algorithms may be used, such as the maximum between-class variance (OTSU) algorithm.

[0092] S22 , vertically stitching the T-line regions of each image in order of concentration to obtain a guide image.

[0093] The T-line ROI areas of the eight colloidal gold immunochromatographic test strip card images captured in the nine multi-exposure time series are stitched together to form an image with eight color levels, which is called the guide image. Figure 5 Middle (B).

[0094] S23, acquiring grayscale value changes of N / X1 pixels in each T-line region in the guide image to obtain a grayscale value matrix of N×P pixels.

[0095] In the guide image, 5 pixels are obtained in each T-line area, and a total of 40 pixels are selected. The changes in the pixel values ​​of the 40 pixels under 9 multi-exposure time sequences are grayscale value matrices, which will be used to fit the camera response function curve. For the pixel value range (Zmax-Zmin) = 255, and each test paper card has 9 multi-exposure photos, select N = 40 > [255 / (9-1) = 32] pixels, then in the 8 T-line ROI areas, select Q = 40 / 8 = 5 pixels each, such as Figure 5 As shown in (C).

[0096] Construct a weight function to optimize the camera response function curve, and the form of the weight function needs to be determined first.

[0097] The standard concentration quantitative curve in the field of immunochromatography often presents an S-shaped characteristic. In the middle of this quantitative curve, there is a linear relationship between the measured pixel value and the concentration of the target substance, while at both ends it enters a plateau period where the correlation becomes less obvious. Therefore, the strategy of piecewise weight function is adopted to more finely reflect the characteristics of each concentration segment, determine the boundary of the linear and nonlinear intervals according to the linear range of the standard concentration quantitative curve, and set the weight coefficient to 1 in the linear interval, thereby retaining the linear characteristics of the original signal.

[0098] The multi-exposure image set is taken at 9 different exposure times and is calculated according to the formula:

[0099]

[0100] Therefore, at least 32 pixels need to be selected to meet the minimum requirement of the number of pixels required for fitting the camera response function curve. In this embodiment, 40 pixels are selected to fit the camera response function curve of the colloidal gold immunochromatography microscopic image acquisition device.

[0101] If 5 pixels with different grayscale values ​​are selected from the image, the digital pixel value matrix of the 5 pixels in the image under 9 different exposure times is obtained, and the corresponding camera response function g curve and irradiance Ei are derived based on this.

[0102] Let X ij is the exposure of the i-th point in the j-th exposure. When the ambient light remains unchanged, this exposure is proportional to the product of the irradiance of this point and the exposure time, where α is the proportional factor:

[0103] X ij =αE i ×Δt j

[0104] Z ij =f(X ij )=f(αE i ×Δt j )

[0105] lnf -1 (Z ij )=lnαE i +lnΔt j

[0106] g(Z ij )=lnαE i +lnΔt j

[0107] Then suppose: i =lnαE i , then λ i is the exposure offset of the pixel point, and the goal is to solve the optimization problem:

[0108]

[0109] Considering the different contribution values ​​of different pixels and emphasizing the weight in the middle of the curve, a weight function needs to be introduced:

[0110]

[0111] Therefore, it is known that Z ij and Δt j , use the least squares method to solve the function g and the exposure offset λ i .

[0112] This embodiment analyzes the application of the identity function, the hat function and the Gaussian function in the camera response function curve fitting, and obtains the following Fig. 9 The five pixels shown have different exposure offsets calculated under different weight functions and Fig.10 Three response function curves g are shown.

[0113] Among them, the identity function form is:

[0114] w1(z)=1,z∈[0,255]

[0115] The functional form of the hat function is:

[0116]

[0117] The functional form of the Gaussian function is:

[0118]

[0119] Among them, parameter a controls the shape of the Gaussian function. A larger a value leads to a wider range of pixels with a weight value of 1 in the weight function. Parameter b controls the width of the Gaussian function. A larger b value makes the distribution range of pixels with a weight value greater than 0 narrower.

[0120] pass Fig.10 The experimental results are obtained, and the residuals of the camera response function curve obtained under three different weight functions, namely, the identity function, the hat function and the Gaussian function, are calculated, which are 6.24503, 0.27919 and 0.19764 respectively. The experimental results show that the fitting degree of the CRF curve under the Gaussian function is better than that of the curve fitted under the identity function and the hat function.

[0121] Furthermore, different (a, b) parameter combinations of the Gaussian function will fit different CRF results, and different fitting results will directly affect the pixel value of each pixel in the HDR image. Therefore, this embodiment establishes a parameter selection algorithm to optimize the selection of a and b of the Gaussian function. The a and b of the Gaussian function are respectively taken in a certain interval, where a is between [2, 50], and a is an even number, and b is between [1, 4]. Because when a = 50 and b = 4, the Gaussian function has set the weights of most pixel values ​​to 1, which is similar to the identity function.

[0122] The MAPE and RMSE values ​​between the pixel value of the T line in the HDR image corrected by the CRF curve fitted with different parameters a and b and the actual T line printed grayscale value are calculated. By comparing the MAPE and RMSE results of 36 test strips under different a and b parameter combinations, the optimal CRF curve is determined, so that the pixel value of the T line in the corrected HDR image is closer to the actual printed grayscale value. The comparison results are shown in Figure 2. Fig.11 shown.

[0123] The experimental results show that when the parameter combination is (a=10, b=4), the obtained MAPE and RMSE values ​​reach the minimum, which are 41.16% and 11.42 respectively. This shows that under this parameter combination, the CRF curve fitted by the Gaussian function can effectively correct the HDR image so that its T-line pixel value can be closest to the actual printed grayscale value.

[0124] According to the optimal function form of the Gaussian function, we can draw the following Fig.12From the weight function diagram shown, it can be found that the weight value begins to turn when the pixel value is 50 and 200, which means that these two pixel values ​​approximately mark the end of the low brightness area and the beginning of the high brightness area in the response function.

[0125] So we construct a piecewise function that behaves differently in low-brightness areas and high-brightness areas. The functional form of the piecewise function is:

[0126]

[0127] Wherein, the function forms of f1(z) and f2(z) include linear function, quadratic polynomial function, exponential function or logarithmic function, and z1 and z2 are segmentation points;

[0128] After determining the form of the segment weight function, it is necessary to solve the segment weight function. First, set the values ​​of the initial segment points z1 and z2 to obtain the first segment weight function; in this embodiment, set the value of segment point z1 to 50 and the value of segment point z2 to 200, and then select the appropriate functions of f1(z) and f2(z) on this basis.

[0129] Select linear, quadratic polynomial, exponential, logarithmic and other functions as candidates for f1(z) and f2(z), and obtain the current camera response function according to the current segmented weight function, i.e., the first segmented weight function, combined with the gray value matrix. The average absolute percentage error MAPE and root mean square error RMSE values ​​between the pixel value of the T line in the HDR image and the actual printed gray value of the T line are corrected by the camera response function curve. Replace the function type of f1(z) and f2(z), update the first segmented weight function, and thus update the camera response function, update the calculated MAPE and RMSE values, and use MAPE and RMSE as the objective function to solve f1(z) and f2(z) that minimize MAPE and RMSE as the final expression of the segmented function. Find the optimal segmented function form and compare the results with the optimal function form of the Gaussian function.

[0130] Specifically, the pixel values ​​of 36 test strips divided into 9 groups at 9 exposure times are shown in Fig.13Middle (A). The T-line pixel values ​​of test strips 1-4 (group 1, and so on) were taken at 31.25ms, the T-line pixel values ​​of test strips 5-8 were taken at 15.625ms, the T-line pixel values ​​of test strips 9-12 were taken at 7.812ms, the T-line pixel values ​​of test strips 13-16 were taken at 3.906ms, the T-line pixel values ​​of test strips 17-20 were taken at 1.953ms, the T-line pixel values ​​of test strips 21-24 were taken at 0.977ms, the T-line pixel values ​​of test strips 25-28 were taken at 0.488ms, the T-line pixel values ​​of test strips 29-32 were taken at 0.244ms, and the T-line pixel values ​​of test strips 33-36 were taken at 0.122ms. Fig.13 The slope of the mapping between the pixel value of the T line of each test paper card and the actual printed grayscale value shown in (B) reflects the sensitivity of the camera to the change of grayscale value under a specific exposure time. The group with a higher slope shows the high sensitivity of the camera to the change of grayscale value, while the group with a lower slope shows that the camera is less sensitive to the change of grayscale value in this range.

[0131] The optimal f1(z) and f2(z) function forms are selected by adjusting the piecewise function form through the values ​​of MAPE and RMSE. The results of calculating the MAPE and RMSE values ​​under different piecewise functions are shown in Table 1. The experimental results show that when f1(x) is an exponential function and f2(x) is a linear function, the obtained MAPE and RMSE values ​​reach the minimum, indicating that the piecewise function under this combination can achieve the optimal correction of HDR images. The MAPE of the T-line pixel values ​​extracted from the original image without HDR algorithm processing and the T-line printed grayscale values ​​is calculated, and the result reaches 117%.

[0132] Table 1 Error results of different f1(x) and f2(x) in piecewise function

[0133]

[0134] After updating the expression of the first segment weight function, the value range of the segment points needs to be optimized. The search range of z1 and z2 is related to the performance of the CMOS image sensor used. The search range of segment points z1 and z2 is set to obtain the second segment weight function. In this embodiment, by observing Fig.13 In the T-line pixel value of the digital image of the test paper card, when the printed grayscale of the test paper card is around 50, the slope of the multi-exposure pixel value becomes 0.9, which is close to linear, indicating that the camera response is good at this time, so the search range of z1 is set to 40-70 (grayscale value). For the test paper card with a high printed grayscale value, the slope decreases to 0.74 when the pixel value is around 200, so the search range of z2 is set to 190-230 (grayscale value).

[0135] Perform a cycle on z1 and z2 of the segmented weight function, set the step size to 1, gradually update the second segmented weight function, calculate the MAPE and RMSE values ​​between the pixel value of the T line in the HDR image corrected with different z1 and z2 and the actual printed T line grayscale value, and determine the segmentation point of the target weight function based on the minimum results of MAPE and RMSE. Through the optimization algorithm, it is found that when z1=50 and z2=200, the MAPE of 36 test strips is 36.21% and the RMSE is 11.23, achieving the minimum error. And this result is better than the MAPE=41.16% and RMSE=11.42 calculated by the Gaussian function objective function form. Therefore, the form of the target weight function in this embodiment is as follows:

[0136]

[0137] After obtaining the target segment weight function, the target camera response function can be obtained according to the target segment weight function combined with the gray value matrix.

[0138] The above objective piecewise function is as follows Fig.14 As shown in A, under the above objective piecewise function Fig.12 The multi-exposure image set of the guide image in is fitted to draw the camera response function curve of the image acquisition device, such as Fig.14 As shown in (B).

[0139] Optionally, S6 specifically includes:

[0140] S61, obtaining an average pixel value of a T-line region of the standard immunochromatographic test paper card according to a target camera response function, a multi-exposure image set of a standard immunochromatographic test paper card, and a corresponding exposure time;

[0141] S62. With the average pixel value in S61 as the y-axis and the actual concentration of the standard immunochromatographic test paper card as the x-axis, a four-parameter logistic fit is used to obtain a standard concentration quantitative curve.

[0142] The standard concentration quantitative curve of immunochromatography is fitted using a four-parameter logistic fitting model, which is usually expressed as follows:

[0143]

[0144] Among them, A1 represents the maximum eigenvalue of the curve, that is, the maximum average pixel value of the T line in the measured image; A2 represents the minimum eigenvalue of the curve, that is, the minimum average pixel value of the T line in the measured image; x represents the independent variable, that is, the printed gray value of the T line; x0 is EC50 (half-maximum effect concentration), that is, the printed gray value corresponding to the middle pixel value of the T line in the measured image; p controls the slope of the curve. At the same time, the four-parameter logistic fitting model has an important evaluation parameter, the fitting degree R 2 It reflects the fitting degree of the standard concentration quantitative curve. The closer its value is to 1, the more accurate the fitted curve is. By fitting the experimental data to the four-parameter logistic model, these parameters can be estimated and an S-shaped curve suitable for the data can be obtained.

[0145] The results of 30 test strips were selected to fit the standard concentration quantitative curve, such as Fig.15 As shown in Table 2, the horizontal axis is the grayscale value of the T line of 30 test strips, and the vertical axis is the change of the T line pixel value in the HDR image. By observing the curve, the data points are closely distributed around the fitting curve, and the fitting degree is 0.99909, and the fitting effect is very good. Secondly, the other 6 test strips that did not participate in the fitting of the standard concentration quantitative curve were used to verify the correctness of the curve fitting, and the predicted printing grayscale value was compared with the actual printing grayscale value. As shown in Table 2, the curve fitting degree is indeed very high, and the maximum difference in pixel value is only 2.

[0146] Table 2 Predicted printing grayscale values ​​calculated by HDR algorithm and actual printing grayscale values ​​of 6 test paper cards

[0147]

[0148] After obtaining the standard concentration quantitative curve, it also includes:

[0149] Shooting the paper card to be tested in a multi-exposure time sequence to obtain a multi-exposure image set of the paper card to be tested;

[0150] The actual concentration value of the paper card to be tested is obtained according to the standard concentration quantitative curve, the multi-exposure image set of the paper card to be tested and the corresponding exposure time.

[0151] The effectiveness of the algorithm designed in this scheme was verified by analytical methodology. The standard concentration quantitative curves of the colloidal gold immunochromatography microscopy image acquisition device before and after the HDR algorithm was used were compared. That is, the T-line pixel values ​​of the LDR images of 30 test strips taken under normal exposure (1.953ms) were compared with the T-line pixel values ​​of the 36 HDR images synthesized using the HDR algorithm, and the upper and lower limits of the measurable signal value area S of the two standard concentration quantitative curves were plotted. The results are shown in Figure 2. Fig.16As shown in the figure, before the HDR algorithm, the range of test strips from 5# to 34# can be measured, and after the HDR algorithm, the range is extended to test strips from 4# to 36#, showing the effectiveness of the multi-exposure fusion HDR algorithm based on the target weight function for the extended range of colloidal gold immunochromatography images. In addition, the linear measurement range of the standard concentration quantitative curve has also been significantly extended.

[0152] In order to verify the applicability of the high dynamic range algorithm in this study, it was applied to the Nadacon commercial handheld detector, such as Fig.17 As shown in the figure, the standard concentration quantitative curves before and after the use of the multi-exposure HDR algorithm based on the target weight function are compared, and the handheld detector uses the T / C value as the characteristic value. The experimental results show that the fitting degree of the curve is improved to 0.99907 after using the HDR algorithm, and the linear range is also increased, from 80-250 to 50-250 (gray value).

[0153] On the basis of the above embodiment, further, the influence of different numbers of multi-exposure photo sets on the T-line characteristic value in the HDR composite image is considered. Specifically, 3 low-exposure, 3 high-exposure and 3 different exposures (underexposure, normal, overexposure) are used to compare the results with the HDR composed of the original 9 multi-exposure photo sets. The results show that the effect of the HDR image synthesized using only 3 low-exposure or only 3 high-exposure is very poor, but when one underexposure, normal, and overexposure are used, the same effect can be achieved as 9 multi-exposure images. The results are shown in Table 3. This is because the accuracy of the camera response function curve fitted under the standard grayscale color scale card is high enough, so that when synthesizing the test paper card HDR image, 3 multi-exposure images can achieve a good effect. Therefore, by optimizing the number of exposure times, the speed of immune detection can be improved while ensuring image quality, and the detection of large sample sizes will be more efficient.

[0154] Table 3 Comparison of T-line pixel values ​​of different numbers of multi-exposure images after HDR algorithm

[0155] Based on the above verification results, when the actual concentration value of the paper card to be tested is detected and analyzed, only one low-exposure, one high-exposure and one different-exposure multi-exposure image set can be used to obtain the actual concentration value of the paper card to be tested according to the standard concentration quantitative curve with high precision and large dynamic detection range in this scheme. The number of multi-exposure images obtained is reduced, the detection speed and accuracy are improved, and the range of analysis and detection of colloidal gold immunochromatographic test paper cards is expanded.

[0156] Although the above scheme takes colloidal gold immunochromatographic test paper card as an example, this method is also applicable to fluorescent immunochromatographic test paper card and can effectively improve its detection performance.

[0157] In the process of restoring the camera response function of the image acquisition system, the embodiment of the present invention uses a piecewise weight function to adjust the weights of images with different exposures, accurately adjusts the fusion of multiple exposure images, and reduces the influence of over-exposure and under-exposure areas in the image on the fitting accuracy of the camera response function, so that the standard quantitative curve obtained by fitting fits the characteristics of the immunochromatographic S-curve, obtains a wider brightness range at different exposure levels, effectively overcomes the limitation of loss of highlight or dark details in a single exposure image, and can significantly improve the accuracy of HDR image synthesis. It effectively reduces the influence of image noise and quantization errors, makes the synthesized immunochromatographic image closer to the real scene, and makes the analysis and detection results based on the immunochromatographic test paper card more accurate. It solves the technical problem that the dynamic range of photos in traditional HDR technology is limited, resulting in the immunochromatographic detector based on the image imaging principle having a narrow detection range, slow detection speed and low detection accuracy, and improves the detection speed and detection accuracy, and expands the scope of analysis and detection of immunochromatographic test paper cards.

[0158] Embodiment 2

[0159] On the basis of the above embodiments, the present invention further provides an analysis and detection device based on a weight function and multi-exposure fusion, characterized in that it is used to perform the analysis and detection method as described in any one of the embodiments, including:

[0160] An image acquisition module is used to photograph the paper card to be tested in a multi-exposure time sequence to obtain a multi-exposure image set of the paper card to be tested;

[0161] The concentration analysis and detection module is used to obtain the actual concentration value of the paper card to be tested according to the standard concentration quantitative curve, the multi-exposure image set of the paper card to be tested and the corresponding exposure time;

[0162] Wherein, the method for generating the standard concentration quantitative curve includes:

[0163] Shooting P multi-exposure time sequences of X standard immunochromatographic test paper cards of different concentrations to obtain a multi-exposure image set; dividing the multi-exposure image set into a first multi-exposure image set including X1 standard immunochromatographic test paper card images and a second multi-exposure image set including X2 standard immunochromatographic test paper card images; wherein P≥3, X=X1+X2;

[0164] Acquire gray value changes of N pixels in the T-line region of each image in the first multi-exposure image set to obtain a gray value matrix of N×P pixels;

[0165] Construct the following piecewise weight function:

[0166]

[0167] Wherein, the function forms of f1(z) and f2(z) include linear function, quadratic polynomial function, exponential function or logarithmic function, and z1 and z2 are segmentation points;

[0168] Determine the function form of f1(z) and f2(z) and the segmentation points z1 and z2 that minimize the error between the average pixel value of the T-line area and the true pixel value, and obtain the target segmentation weight function; the calculation method of the average pixel value of the T-line area includes: obtaining a camera response function according to the segmentation weight function combined with the gray value matrix; obtaining an HDR image according to the camera response function, the second multi-exposure image set and the corresponding exposure time; and performing tone mapping on the HDR image to obtain the average pixel value of the T-line area;

[0169] The target camera response function is obtained according to the target segment weight function combined with the gray value matrix;

[0170] According to the target camera response function, the multi-exposure image set of the standard immunochromatographic test strip card and the corresponding exposure time, the average pixel value of the T-line area of ​​the standard immunochromatographic test strip card is obtained, and the average pixel value of the T-line area of ​​the standard immunochromatographic test strip card is fitted with the corresponding concentration to generate a standard concentration quantitative curve.

[0171] An analysis and detection device based on a weight function and multi-exposure fusion provided in an embodiment of the present invention is used to execute an analysis and detection method based on a weight function and multi-exposure fusion provided in any embodiment 1 of the present invention, and has corresponding functional modules and beneficial effects.

[0172] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An analysis and detection method based on weight function and multi-exposure fusion, characterized in that: include: Shooting the paper card to be tested in a multi-exposure time sequence to obtain a multi-exposure image set of the paper card to be tested; According to the standard concentration quantitative curve, the multi-exposure image set of the paper card to be tested and the corresponding exposure time, the actual concentration value of the paper card to be tested is obtained; The method for generating the standard concentration quantitative curve comprises: S1, photographing P multi-exposure time sequences of X standard immunochromatographic test paper cards of different concentrations to obtain a multi-exposure image set; dividing the multi-exposure image set into a first multi-exposure image set including X1 standard immunochromatographic test paper card images and a second multi-exposure image set including X2 standard immunochromatographic test paper card images; wherein P≥3, X=X1+X2; S2. Obtain grayscale value changes of N pixels in the T-line area of ​​each image in the first multi-exposure image set to obtain a grayscale value matrix of N×P pixels; S3. Construct the following piecewise weight function: Wherein, the function forms of f1(z) and f2(z) include linear function, quadratic polynomial function, exponential function or logarithmic function, and z1 and z2 are segmentation points; S4, determining the function forms of f1(z) and f2(z) and the segmentation points z1 and z2 that minimize the error between the average pixel value of the T-line area and the true pixel value, and obtaining a target segmentation weight function; the calculation method of the average pixel value of the T-line area includes: obtaining a camera response function according to the segmentation weight function combined with a gray value matrix; obtaining an HDR image according to the camera response function, a second multi-exposure image set and a corresponding exposure time; performing tone mapping on the HDR image to obtain an average pixel value of the T-line area; S5, obtaining a target camera response function according to the target segment weight function combined with the gray value matrix; S6. According to the target camera response function, the multi-exposure image set of the standard immunochromatographic test paper card and the corresponding exposure time, the average pixel value of the T-line area of ​​the standard immunochromatographic test paper card is obtained, and the average pixel value of the T-line area of ​​the standard immunochromatographic test paper card is fitted with the corresponding concentration to generate a standard concentration quantitative curve.

2. The analysis and detection method according to claim 1, characterized in that: The standard immunochromatographic test paper cards have different T line concentrations and the same C line concentrations.

3. The analysis and detection method according to claim 1, characterized in that: S2 specifically includes: S21, extracting a T-line region of each image in the first multi-exposure image set using an image segmentation algorithm; S22, vertically stitching the T-line regions of each image in order of concentration to obtain a guide image; S23, acquiring grayscale value changes of N / X1 pixels in each T-line region in the guide image to obtain a grayscale value matrix of N×P pixels.

4. The analysis and detection method according to claim 1, characterized in that: The error between the average pixel value and the true pixel value is the root mean square error RMSE and the mean absolute percentage error MAPE of the average pixel value of the T-line area and the true pixel value of the standard immunochromatographic test paper card.

5. The analysis and detection method according to claim 1, characterized in that: S6 specifically includes: S61, obtaining an average pixel value of a T-line region of the standard immunochromatographic test paper card according to a target camera response function, a multi-exposure image set of a standard immunochromatographic test paper card, and a corresponding exposure time; S62. With the average pixel value in S61 as the y-axis and the actual concentration of the standard immunochromatographic test paper card as the x-axis, a four-parameter logistic fit is used to obtain a standard concentration quantitative curve.

6. The analysis and detection method according to claim 1, characterized in that: The standard immunochromatographic test paper cards include colloidal gold immunochromatographic cards and fluorescent immunochromatographic cards.

7. The analysis and detection method according to claim 1, characterized in that: The multi-exposure image set covers underexposure to overexposure in different exposure time sequences.

8. The analysis and detection method according to claim 1, characterized in that: After obtaining the multi-exposure image set, the method further includes: preprocessing the multi-exposure image set using a median filtering algorithm, a Gaussian filtering algorithm, a bilateral filtering algorithm or a guided filtering algorithm.

9. An analysis and detection device based on weight function and multi-exposure fusion, characterized in that: Used to perform the analysis and detection method according to any one of claims 1 to 8, comprising: An image acquisition module is used to photograph the paper card to be tested in a multi-exposure time sequence to obtain a multi-exposure image set of the paper card to be tested; The concentration analysis and detection module is used to obtain the actual concentration value of the paper card to be tested according to the standard concentration quantitative curve, the multi-exposure image set of the paper card to be tested and the corresponding exposure time; Wherein, the method for generating the standard concentration quantitative curve includes: Shooting P multi-exposure time sequences of X standard immunochromatographic test paper cards of different concentrations to obtain a multi-exposure image set; dividing the multi-exposure image set into a first multi-exposure image set including X1 standard immunochromatographic test paper card images and a second multi-exposure image set including X2 standard immunochromatographic test paper card images; wherein P≥3, X=X1+X2; Acquire gray value changes of N pixels in the T-line region of each image in the first multi-exposure image set to obtain a gray value matrix of N×P pixels; Construct the following piecewise weight function: Wherein, the function forms of f1(z) and f2(z) include linear function, quadratic polynomial function, exponential function or logarithmic function, and z1 and z2 are segmentation points; Determine the function form of f1(z) and f2(z) and the segmentation points z1 and z2 that minimize the error between the average pixel value of the T-line area and the true pixel value, and obtain the target segmentation weight function; the calculation method of the average pixel value of the T-line area includes: obtaining a camera response function according to the segmentation weight function combined with the gray value matrix; obtaining an HDR image according to the camera response function, the second multi-exposure image set and the corresponding exposure time; and performing tone mapping on the HDR image to obtain the average pixel value of the T-line area; The target camera response function is obtained according to the target segment weight function combined with the gray value matrix; According to the target camera response function, the multi-exposure image set of the standard immunochromatographic test strip card and the corresponding exposure time, the average pixel value of the T-line area of ​​the standard immunochromatographic test strip card is obtained, and the average pixel value of the T-line area of ​​the standard immunochromatographic test strip card is fitted with the corresponding concentration to generate a standard concentration quantitative curve.

Images