A colloidal gold reagent strip detection method and system based on deep learning

Abstract

The invention discloses a colloidal gold reagent strip detection method and system based on deep learning, and relates to the technical field of colloidal gold reagent strip detection. The acquired image to be detected is transmitted to a trained deep learning neural network model for image segmentation to obtain a standard color card image, a colloidal gold reagent strip image and a two-dimensional code image carrying standard image parameters; the two-dimensional code image is parsed to obtain a standard grayscale value, the standard color card is parsed to obtain a read grayscale value, and a grayscale value correction curve of the read grayscale value is fitted based on the standard grayscale value; the colloidal gold reagent strip image is parsed to obtain a color grayscale value, the color grayscale value is corrected by using the grayscale value correction curve to obtain a corrected color grayscale value, the corrected color grayscale value is matched with a preset biological calculation algorithm to obtain a detection result of the colloidal gold reagent strip; the detection method of the present application improves the accuracy of the detection result of the colloidal gold reagent strip image.

Description

Technical Field

[0001] The present invention relates to the technical field of colloidal gold reagent strip detection, and in particular to a colloidal gold reagent strip detection method and system based on deep learning. Background Art

[0002] With the improvement of the quality of life, people are more and more concerned about their physical health; and the traditional way to obtain physical health indicators is to go to the hospital for various health indicators. This method is not only costly, but also requires queuing up at the hospital, wasting time, and occupying public medical resources; therefore, users are in urgent need of a solution that can do simple health tests at home. However, the existing colloidal gold test strip health test applications based on smartphones have serious deficiencies: due to differences in mobile phone models, the test images obtained by the cloud server are different, which leads to large detection errors of the colloidal gold test strips; at the same time, due to differences in ambient light intensity, the test images obtained by the cloud server are also different, resulting in large detection errors of the colloidal gold test strips. Summary of the invention

[0003] Based on the technical problems existing in the background technology, the present invention proposes a colloidal gold reagent strip detection method and system based on deep learning, which improves the detection results of colloidal gold reagent strip images under different mobile phone models and different lighting conditions.

[0004] The present invention proposes a colloidal gold reagent strip detection method based on deep learning, comprising:

[0005] The acquired image to be tested is transmitted to the trained deep learning neural network model for image segmentation to obtain a standard color card image, a colloidal gold reagent strip image, and a QR code image carrying image standard parameters;

[0006] The standard grayscale value is obtained by analyzing the two-dimensional code image, and the read grayscale value is obtained by analyzing the standard color card. The grayscale value correction curve of the read grayscale value is fitted based on the standard grayscale value;

[0007] The image of the colloidal gold reagent strip is analyzed to obtain a color gray value, and the color gray value is corrected using a gray value correction curve to obtain a corrected color gray value;

[0008] According to the test items, the corrected color grayscale value is matched with the preset biological calculation algorithm to obtain the test result of the colloidal gold reagent strip.

[0009] Furthermore, the acquired image to be detected is transmitted to the trained deep learning neural network model for image segmentation, and the standard color card image, the colloidal gold reagent strip image and the QR code image carrying the image standard parameters are obtained, including:

[0010] The acquired image to be detected is marked and segmented to obtain a two-dimensional code image, a standard color card image and a colloidal gold reagent strip image respectively;

[0011] Analyze the QR code image to obtain the standard grayscale value of each grayscale area of ​​the standard color card under the standard lighting environment and the biological algorithm of the current detection project;

[0012] Analyze the standard color card image to obtain the read grayscale value of each grayscale area of ​​the current image to be detected;

[0013] The colloidal gold reagent strip image is analyzed to obtain the grayscale value of each color development area of ​​the current image to be detected.

[0014] Furthermore, the training steps of the deep learning neural network model are:

[0015] Acquire multiple original training image data sets, segment each image in the original training image data sets, obtain a two-dimensional code image, a standard color card image, and a colloidal gold reagent strip image carrying image standard parameters, and mark the two-dimensional code image, the standard color card image, and the colloidal gold reagent strip image;

[0016] Determining a training target for training the deep learning neural network model based on the original training image data set;

[0017] The original training image data in the original training image data set is input into the first layer of the deep learning neural network model, and the deep learning neural network model is trained according to the training objective.

[0018] Furthermore, the deep learning neural network model includes 7 3×3 convolutional layers, the output end of the previous convolutional layer is connected to the input end of the next convolutional layer, and PRelu is used as the activation function.

[0019] Furthermore, in obtaining a plurality of original training image data sets, including:

[0020] Get original training images of multiple phone models and multiple light intensities;

[0021] Each original training image is rotated at different angles to obtain an original training image dataset.

[0022] A colloidal gold reagent strip detection system based on deep learning, comprising a segmentation and analysis module, an image fitting module, a correction module and a matching output module;

[0023] The segmentation and analysis module is used to transmit the acquired image to be detected to the trained deep learning neural network model for image segmentation, and obtain the standard color card image, the colloidal gold reagent strip image and the QR code image carrying the image standard parameters;

[0024] The image fitting module is used to analyze the QR code image to obtain the standard grayscale value, analyze the standard color card to obtain the read grayscale value, and fit the grayscale correction curve of the read grayscale value based on the standard grayscale value;

[0025] The correction module is used to analyze the colloidal gold reagent strip image to obtain the color gray value, and correct the color gray value using the gray value correction curve to obtain the corrected color gray value;

[0026] The matching output module is used to match the corrected color gray value with a preset biological calculation algorithm according to the detection item to obtain the detection result of the colloidal gold reagent strip.

[0027] Furthermore, the segmentation and analysis module includes a segmentation module, a two-dimensional code image analysis module, a standard color card image analysis module and a colloidal gold reagent strip image analysis module;

[0028] The segmentation module is used to mark and segment the acquired image to be detected, and obtain a two-dimensional code image, a standard color card image and a colloidal gold reagent strip image respectively;

[0029] The two-dimensional code image analysis module is used to analyze the two-dimensional code image to obtain the standard grayscale value of each grayscale area of ​​the standard color card under the standard lighting environment and the biological algorithm of the current detection project;

[0030] The standard color card image analysis module is used to analyze the standard color card image to obtain the read grayscale value of each grayscale area of ​​the current image to be detected;

[0031] The colloidal gold reagent strip image analysis module is used to analyze the colloidal gold reagent strip image to obtain the color gray value of each color development area of ​​the current image to be detected.

[0032] The advantages of the colloidal gold reagent strip detection method and system based on deep learning provided by the present invention are: the colloidal gold reagent strip detection method and system based on deep learning provided in the structure of the present invention, based on the standard color card and the two-dimensional code carrying the standard image parameters, correct the detection and recognition results of the colloidal gold reagent strip image obtained on different mobile phone models, which well alleviates the difference in photographed images of different models of mobile phones, and ultimately alleviates the difference in detection results on different mobile phone models; at the same time, based on the standard color card and the two-dimensional code carrying the standard image parameters, the results of the colloidal gold reagent strip image under ambient light conditions of different intensities are corrected, which well alleviates the detection error introduced by the intensity of the ambient light. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a structural schematic diagram of the present invention; DETAILED DESCRIPTION

[0034] Below, the technical solution of the present invention is described in detail through specific embodiments. Many specific details are set forth in the following description to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific implementation disclosed below.

[0035] Reference Figure 1 , a colloidal gold reagent strip detection method based on deep learning proposed by the present invention comprises steps S1 to S5:

[0036] S1: The acquired image to be detected is transmitted to the trained deep learning neural network model for image segmentation to obtain a standard color card image, a colloidal gold reagent strip image, and a QR code image carrying image standard parameters;

[0037] Each colloidal gold reagent strip is matched with a standard color card and a QR code carrying standard image parameters before leaving the factory. The user will first place the QR code, standard color card and colloidal gold reagent strip on the same plane, and use the camera on the mobile phone to obtain an image to be tested containing the above three, and obtain an image to be tested containing the QR code, standard color card and colloidal gold reagent strip. Then, the three images in the acquired image to be tested will be marked and segmented through a deep learning neural network model to obtain a QR code image, a standard color card image and a colloidal gold reagent strip image, so as to obtain the test result corresponding to the colloidal gold reagent strip image.

[0038] S2: parsing the QR code image to obtain a standard grayscale value, parsing the standard color card to obtain a read grayscale value, and fitting a grayscale correction curve of the read grayscale value based on the standard grayscale value;

[0039] S3: analyzing the colloidal gold reagent strip image to obtain a color gray value, and correcting the color gray value using a gray value correction curve to obtain a corrected color gray value;

[0040] S4: According to the detection items, the corrected grayscale value is matched with a preset biological calculation algorithm to obtain the detection result of the colloidal gold reagent strip.

[0041] Among them, the biological calculation algorithm is an existing biological algorithm parsed from the QR code image. Because the colloidal gold reagent strip can be used for tests with different needs, that is, the test items can be gastric function tests, lung function tests, new coronavirus tests, etc., in order to more accurately quantify the colloidal gold reagent strip, each test item corresponds to a biological calculation algorithm. In actual application, the colloidal gold reagent strip will match the preset biological calculation algorithm according to the test item and generate a test report for users to view, so as to improve the accuracy of the test results of different test items.

[0042] S5: The detection result of the colloidal gold reagent image is transmitted to the mobile phone for the user to view.

[0043] Generally, the detection parameters of the colloidal gold reagent image are generated into a detection report for the convenience of users to view.

[0044] If there is a result reference range for the items generally tested, the concentration or content of the colloidal gold reagent image obtained can be compared with the result reference range, and the normal (within the result reference range), high (higher than the result reference range) and low (lower than the result reference range) results can also be displayed in the final search report and pushed to the user.

[0045] At the same time, the search report can also include suggestions. For example, when conducting a gastric function test, if the test results are abnormal (too low or too high), the user can be advised to reduce the intake of irritating foods, seek medical attention, etc., so that the user can not only accurately obtain the results of the test, but also intuitively get the direction of improvement.

[0046] There is a one-to-one correspondence between the detection items and the biological calculation algorithms. Therefore, the biological calculation algorithm of the detection item parsed through the QR code is directly used as the pre-set biological calculation algorithm to obtain the detection result of the colloidal gold reagent strip corresponding to the colloidal gold reagent strip image data, which simplifies the user's operation steps and improves the user experience.

[0047] Through steps S1 to S5, the present application uses a deep learning neural network model to mark and segment the image to be detected, which is convenient for correcting the colloidal gold reagent strip image through a grayscale value correction curve, and can output relatively accurate detection concentration and content of the detection items. It can avoid the defect of inaccurate detection concentration and content of the detection items finally obtained due to different mobile phone parameters and models when using traditional mobile phones for detection. Therefore, the present application can be suitable for users' needs for different actual detection scenarios, has high detection accuracy, and is widely used.

[0048] The colloidal gold reagent strip detection of the present application can be applied not only to mobile device systems (Android system, iOS system, etc.), but also to computer systems (Windows system, Linux system, etc.); not only to client applications, but also to web front-ends.

[0049] Further, in step S1: the acquired image to be detected is transmitted to the trained deep learning neural network model for image segmentation to obtain a standard color card image, a colloidal gold reagent strip image and a two-dimensional code image carrying image standard parameters, including steps S11 to S14:

[0050] S11: marking and segmenting the acquired image to be detected to obtain a two-dimensional code image, a standard color card image, and a colloidal gold reagent strip image;

[0051] The images to be detected are marked and segmented through a deep learning neural network model.

[0052] S12: parsing the two-dimensional code image to obtain the standard grayscale value of each grayscale area of ​​the standard color card under a standard lighting environment and the biological algorithm of the current detection item;

[0053] The standard grayscale value corresponding to the QR code is the standard grayscale value of each grayscale area of ​​each batch of standard color cards tested under a standard light box environment, and the standard grayscale value is saved in the QR code. For example, if 11 grayscale areas are pre-set, 11 corresponding standard grayscale values ​​are obtained.

[0054] S13: Analyze the standard color card image to obtain the read grayscale value of each grayscale area of ​​the current image to be detected;

[0055] The standard color card is used as a benchmark for comparing the grayscale values ​​of the colloidal gold reagent strip image when it is displayed. Due to the image processing function of the mobile phone camera itself, the standard color card has certain distortion. Therefore, this application also sets up a QR code image to correct the processing effect of different mobile phone cameras on the standard color card image, so that the colloidal gold reagent strip image in this application can be suitable for image acquisition of mobile phones of different specifications, and the final detection result corresponding to the colloidal gold reagent strip image will not have a large error defect due to certain preprocessing of the image to be detected by different mobile phones.

[0056] S14: Analyze to obtain the grayscale value of each color-developed area of ​​the current image to be detected.

[0057] The colloidal gold reagent strip image and the standard color card image are two images taken with the same camera and the same parameters on the same mobile phone. Therefore, the QR code image, the standard color card image and the colloidal gold reagent strip image are three images taken with the same mobile phone camera and the same camera parameters. Therefore, the image processing process of the camera itself for the above three images is consistent. Therefore, the three images are comparable after being marked and segmented in the deep learning neural network model.

[0058] Through steps S11 to S14, the acquisition and segmentation of the image to be detected are realized, which facilitates the further processing of the image to be detected.

[0059] Further, in step S2: the two-dimensional code image is parsed to obtain a standard grayscale value, the standard color card is parsed to obtain a read grayscale value, and a grayscale correction curve of the read grayscale value is fitted based on the standard grayscale value, including steps S21 to S22:

[0060] S21 analyzes the QR code image to obtain a standard grayscale value, and analyzes the standard color card to obtain a read grayscale value;

[0061] The parsed QR code image includes the standard grayscale value and the biological algorithm of the current detection item. The parsed standard color card image includes the read grayscale value. In theory, the standard grayscale value is consistent with the read grayscale value. However, since different mobile phone cameras have different image processing processes for images, there is a certain image difference between the read grayscale value and the standard grayscale value.

[0062] S22: Based on the standard gray value, a gray value correction curve of the read gray value is fitted.

[0063] The gray value correction curve can be fitted by the following formula:

[0064] Y=A×X 2 +B×X+C

[0065] Among them, X is the original grayscale value, Y is the standard grayscale value, and A, B, and C are parameter values.

[0066] In the fitting process, the least square method is used, with the standard gray value as the Y axis (reference) and the read gray value as the X axis, and a gray value correction curve with the standard gray value as the reference and the read gray value as the fitting.

[0067] During the fitting calculation process, the parameter values ​​of A, B, and C are calculated based on the standard grayscale values ​​of the 11 color levels retained in the QR code (the grayscale values ​​measured in a standard light source environment before leaving the factory) and the read grayscale values ​​of the 11 color levels calculated from the image taken by the mobile terminal (such as a mobile phone), and the above-mentioned correction curve is fitted.

[0068] According to steps S21 to S22, since the shooting environment and shooting parameters of the colloidal gold reagent strip image and the standard color card image are consistent, the fitting curve of the read grayscale value and the standard grayscale value can be used to correct the color grayscale value to obtain the corrected color grayscale value; it should be understood that the corrected color grayscale value obtained at this time is a correction of the color grayscale value directly uploaded by the mobile phone, which avoids the image difference between the standard color card image and the colloidal gold reagent strip image obtained by different mobile phone models under different lighting environments, and realizes the correction of the colloidal gold reagent strip test results under different mobile phone models and different lighting environments, alleviates the detection error caused by the intensity of the lighting environment, and can adapt to various detection scenarios of users.

[0069] In step S3: the image of the colloidal gold reagent strip is analyzed to obtain a color gray value, and the color gray value is corrected using a gray value correction curve to obtain a corrected color gray value, including steps S31 to S32:

[0070] S31: analyzing the colloidal gold reagent strip image to obtain a color gray value;

[0071] S32: Bringing the color gray value into the gray value correction curve for correction to obtain the corrected color gray value.

[0072] Through steps S3 to S32, the corrected color grayscale value is obtained, which realizes the influence of image preprocessing on the image to be detected when different mobile phones upload the image to be detected.

[0073] Furthermore, the training steps S51 to S53 of the deep learning neural network model are:

[0074] S51: Acquire multiple original training image data sets, and mark each image in the original training image data sets.

[0075] Mark and segment the QR code image, standard color card image, and colloidal gold reagent strip image of each image respectively;

[0076] The method for obtaining multiple original training image data sets includes: obtaining original training images of multiple mobile phone models and multiple light intensities; and rotating each original training image at different angles to obtain the original training image data sets.

[0077] Different models of mobile phones are used to acquire original training images in different environments.

[0078] When rotating each original training image at different angles, each original training image can be rotated 45 degrees, 90 degrees, 135 degrees, 180 degrees, 225 degrees, 270 degrees, and 315 degrees, as well as flipped left and right, to construct a larger training sample set, and back up each of these rotated images, and mark and segment one of them.

[0079] The original training image dataset includes unlabeled images and corresponding labeled images.

[0080] S52: Determine a training target for training the deep learning neural network model based on the original training image data set;

[0081] The deep learning neural network model includes 7 3×3 convolutional layers, the output end of the previous convolutional layer is connected to the input end of the next convolutional layer, and PRelu is used as the activation function.

[0082] The convolution kernels included in the convolution layer of the deep learning neural network model are two-dimensional matrices. The effect of 7 layers of 3×3 convolution kernels on the image is equivalent to that of one layer of 15×15 convolution kernels. Assume that the image resolution of the input image to be processed is 100×100, and the convolution operation does not change the size. For a convolution layer including m 15×15 convolution kernels, 100×100×15×15×m multiplication operations are required when the image to be processed is convolved through the convolution layer. For 7 layers of convolution layers each including m 3×3 convolution kernels, 100×100×3×3×m×7 multiplication operations are required when the image to be processed is convolved through these 7 layers of convolution layers. The same parts are reduced to 15×15=225 and 3×3×7=63 respectively. Obviously, replacing one layer of 15×15 convolution layer with 7 layers of 3×3 convolution layers can reduce the amount of calculation by nearly 70%. Therefore, in this solution, using a convolution kernel of size 3×3 for each convolution layer can improve the operation speed of the deep learning neural network model. Moreover, more convolution layers can improve the expression ability of the deep learning neural network model to a certain extent. Therefore, the marking and segmentation of the image to be detected can be achieved.

[0083] Moreover, when setting the activation function for the convolutional layer of the deep learning neural network model, the PRelu function is selected as the activation function. Using the PRelu function as the activation function can achieve faster convergence speed, because the output mean of the PRelu function is closer to 0, and the convergence is faster in the deep learning iterative calculation.

[0084] S53: Input the original training image data in the original training image data set from the first layer of the deep learning neural network model, and train the deep learning neural network model according to the training objective.

[0085] The deep learning neural network model is trained through labeled images, and the segmented image obtained after the unlabeled image passes through the deep learning neural network model is compared with the corresponding labeled image. The deep learning neural network model can be corrected to improve the image processing accuracy of the deep learning neural network model.

[0086] When the actual image to be detected is processed by the trained deep learning neural network model, the image to be detected will be marked and segmented, which can reduce the influence of the image to be detected due to tilt, shooting angle, etc., and obtain the segmented image to be detected, so as to achieve a more accurate detection result through the correction output of the colloidal gold reagent strip image.

[0087] A colloidal gold reagent strip detection system based on deep learning, comprising a segmentation and analysis module, an image fitting module, a correction module and a matching output module;

[0088] The segmentation and analysis module is used to transmit the acquired image to be detected to the trained deep learning neural network model for image segmentation, and obtain the standard color card image, the colloidal gold reagent strip image and the QR code image carrying the image standard parameters;

[0089] The image fitting module is used to analyze the QR code image to obtain the standard grayscale value, analyze the standard color card to obtain the read grayscale value, and fit the grayscale correction curve of the read grayscale value based on the standard grayscale value;

[0090] The correction module is used to analyze the colloidal gold reagent strip image to obtain the color gray value, and correct the color gray value using the gray value correction curve to obtain the corrected color gray value;

[0091] The matching output module is used to match the corrected color gray value with a preset biological calculation algorithm according to the detection item to obtain the detection result of the colloidal gold reagent strip.

[0092] Furthermore, the segmentation and analysis module includes a segmentation module, a two-dimensional code image analysis module, a standard color card image analysis module and a colloidal gold reagent strip image analysis module;

[0093] The segmentation module is used to mark and segment the acquired image to be detected, and obtain a two-dimensional code image, a standard color card image and a colloidal gold reagent strip image respectively;

[0094] The two-dimensional code image analysis module is used to analyze the two-dimensional code image to obtain the standard grayscale value of each grayscale area of ​​the standard color card under the standard lighting environment and the biological algorithm of the current detection project;

[0095] The standard color card image analysis module is used to analyze the standard color card image to obtain the read grayscale value of each grayscale area of ​​the current image to be detected;

[0096] The colloidal gold reagent strip image analysis module is used to analyze the colloidal gold reagent strip image to obtain the color gray value of each color development area of ​​the current image to be detected.

[0097] A computer-readable storage medium stores a plurality of acquisition and classification programs, wherein the plurality of acquisition and classification programs are used to be called by a processor and execute the colloidal gold reagent strip detection method as described above.

[0098] A person of ordinary skill in the art can understand that all or part of the steps of implementing the above method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiment; and the aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk, etc., various media that can store program codes.

[0099] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A colloidal gold reagent strip detection method based on deep learning, characterized in that: include: S1. The acquired image to be detected is transmitted to the trained deep learning neural network model for image segmentation to obtain a standard color card image, a colloidal gold reagent strip image, and a QR code image carrying image standard parameters, specifically including: S11, marking and segmenting the acquired image to be detected to obtain a two-dimensional code image, a standard color card image and a colloidal gold reagent strip image respectively; S12, parsing the two-dimensional code image to obtain the standard grayscale value of each grayscale area of ​​the standard color card under a standard lighting environment and the biological algorithm of the current detection item; S13, analyzing the standard color card image to obtain the read grayscale value of each grayscale area of ​​the current image to be detected; S14, analyzing the colloidal gold reagent strip image to obtain the color grayscale value of each color development area of ​​the current image to be detected; S2, parsing the two-dimensional code image to obtain a standard grayscale value, parsing the standard color card to obtain a read grayscale value, and fitting a grayscale correction curve of the read grayscale value based on the standard grayscale value; S3, analyzing the colloidal gold reagent strip image to obtain a color gray value, and correcting the color gray value using a gray value correction curve to obtain a corrected color gray value; S4. According to the detection items, the corrected grayscale value is matched with a preset biological calculation algorithm to obtain the detection result of the colloidal gold reagent strip.

2. The colloidal gold reagent strip detection method based on deep learning according to claim 1, characterized in that: The training steps of the deep learning neural network model are: Acquire multiple original training image data sets, and label each image in the original training image data sets; Determining a training target for training the deep learning neural network model based on the original training image data set; The original training image data in the original training image data set is input into the first layer of the deep learning neural network model, and the deep learning neural network model is trained according to the training objective.

3. The colloidal gold reagent strip detection method based on deep learning according to claim 2, characterized in that: The deep learning neural network model includes 7 3×3 convolutional layers, the output end of the previous convolutional layer is connected to the input end of the next convolutional layer, and PRelu is used as the activation function.

4. The colloidal gold reagent strip detection method based on deep learning according to claim 3, characterized in that: In obtaining multiple original training image datasets, including: Get original training images of multiple phone models and multiple light intensities; Each original training image is rotated at different angles to obtain an original training image dataset.

5. A colloidal gold reagent strip detection system based on deep learning, characterized in that: It includes a segmentation and analysis module, an image fitting module, a correction module and a matching output module. The segmentation and analysis module includes a segmentation module, a two-dimensional code image analysis module, a standard color card image analysis module and a colloidal gold reagent strip image analysis module; The segmentation and analysis module is used to transmit the acquired image to be detected to the trained deep learning neural network model for image segmentation, and obtain the standard color card image, the colloidal gold reagent strip image and the QR code image carrying the image standard parameters; The image fitting module is used to analyze the QR code image to obtain the standard grayscale value, analyze the standard color card to obtain the read grayscale value, and fit the grayscale correction curve of the read grayscale value based on the standard grayscale value; The correction module is used to analyze the colloidal gold reagent strip image to obtain the color gray value, and correct the color gray value using the gray value correction curve to obtain the corrected color gray value; The matching output module is used to match the corrected color gray value with a preset biological calculation algorithm according to the detection item to obtain the detection result of the colloidal gold reagent strip; The segmentation module is used to mark and segment the acquired image to be detected, and obtain a two-dimensional code image, a standard color card image and a colloidal gold reagent strip image respectively; The two-dimensional code image analysis module is used to analyze the two-dimensional code image to obtain the standard grayscale value of each grayscale area of ​​the standard color card under the standard lighting environment and the biological algorithm of the current detection project; The standard color card image analysis module is used to analyze the standard color card image to obtain the read grayscale value of each grayscale area of ​​the current image to be detected; The colloidal gold reagent strip image analysis module is used to analyze the colloidal gold reagent strip image to obtain the color gray value of each color development area of ​​the current image to be detected.

Images