A digital identification method suitable for colloidal gold detection card detection

Abstract

The application belongs to the field of colloidal gold detection, and specifically discloses a digital identification method suitable for colloidal gold detection card detection, and the specific steps are as follows: S1, image acquisition; S2, image preprocessing; S3, region detection; S4, feature extraction; S5, data comparison and analysis; S6, system integration; by integrating a complete application program, after the colloidal gold detection card is shot by a mobile phone or an iPad, the image is uploaded to a cloud server for processing, one or more regions containing colloidal gold detection card C and T lines are identified, the length, width and color value of the C and T lines are read and calculated, the extracted C and T line values are compared with a standard dictionary, the content of a target substance in a sample is judged, and the data is uploaded to a food safety big data platform, without relying on professional detection instruments, the method is simple to operate, and is convenient for food enterprises, farmers' markets, food small workshops and families to perform self-checking screening on food, and food safety is protected.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of colloidal gold detection, and particularly relates to a digital recognition method suitable for colloidal gold detection card detection. BACKGROUND

[0002] The colloidal gold detection card is a tool for rapid detection, and is widely applied to the fields of medicine, food safety and environmental monitoring, and the working principle thereof is based on the combination of the colloidal gold labeled antibody or antigen and the target substance in a sample, and the color change is used to indicate the result.

[0003] The traditional detection and recognition method mainly uses a special detection instrument to read the detection data of the colloidal gold single card, and then a colorimetric method is used to determine the content of the target substance in the sample.

[0004] However, the traditional detection and recognition method needs to rely on professional detection instruments, has high cost, is inconvenient for food enterprises, farmers' markets, food workshops, families and the like to make self-checking screening on food, and is not conducive to the detection and recognition of multi-card and multi-line cards. SUMMARY

[0005] The application aims to provide a digital recognition method suitable for colloidal gold detection card detection, so as to solve the problems in the background.

[0006] To achieve the above-mentioned purpose, the application provides the following technical scheme.

[0007] A digital recognition method suitable for colloidal gold detection card detection, and the specific steps are as follows.

[0008] S1, image acquisition: the colloidal gold detection card is photographed by using a mobile phone or an iPad, and after ensuring that the image is clear and the light is uniform, the image is uploaded to a cloud server for processing;

[0009] S2, image preprocessing: the image is processed by using an image processing technology;

[0010] S3, region detection: the C and T line regions in the image are recognized by using an edge detection algorithm, and then the contours of the C and T lines are extracted by using a contour detection algorithm;

[0011] S4, feature extraction: the length, width and color value of the C and T lines are calculated, and the threshold range of the CT line is set according to a standard dictionary;

[0012] S5, data comparison and analysis: the extracted C and T line values are compared with the standard dictionary, the content of the target substance in the sample is determined, the detection result is recorded, and the data is uploaded to a food safety big data platform;

[0013] S6, System integration: integrate S2-S5 into a complete application, provide a user-friendly interface for users to upload images and view detection results.

[0014] Preferably, the specific steps of image preprocessing in S2 are as follows:

[0015] S2.1, Denoising: use Gaussian filter to smooth the image to reduce noise;

[0016] S2.2, Grayscale: convert color image to grayscale image.

[0017] Preferably, the specific steps of region detection in S3 are as follows:

[0018] Sa3.1, Gradient calculation: use Sobel operator to calculate the gradient amplitude and direction of the image;

[0019] Sa3.2, Non-maximum suppression: non-maximum suppression is performed on the gradient amplitude, and the local maximum is retained, and the non-edge pixels are suppressed;

[0020] Sa3.3, Double threshold processing: set high threshold and low threshold, and divide the edge into strong edge, weak edge and non-edge;

[0021] Sa3.4, Edge connection: connect strong edge and weak edge to form complete edge;

[0022] Sa3.5, Extract contour: use OpenCV's findContours function to extract the contour in the edge image;

[0023] Sa3.6, Draw contour: use drawContours function to draw the extracted contour on the original image.

[0024] Preferably, in the Sa3.2 step, whether to retain the pixel is determined by comparing the current pixel with its adjacent pixels in the gradient direction.

[0025] Preferably, in the Sa3.3 step, the strong edge is the pixel greater than the high threshold, the weak edge is the pixel between the high threshold and the low threshold, and the non-edge is the pixel less than the low threshold.

[0026] Preferably, in the Sa3.4 step, if the weak edge is connected with the strong edge, the weak edge is retained, otherwise the weak edge is suppressed.

[0027] Preferably, in the S3 step, before identifying the region of C and T lines and extracting the contour of C and T lines, OTSU threshold segmentation algorithm is used for segmentation processing of the image, and the specific steps are as follows:

[0028] Sb3.1, Calculate image histogram: Count the number of pixels in each gray level in the image to generate a gray histogram;

[0029] Sb3.2, Calculate total number of pixels in the image: Calculate the total number of all pixels in the image;

[0030] Sb3.3, Calculate the probability of each gray level: Calculate the probability of each gray level;

[0031] Sb3.4, Calculate the mean of foreground and background: Calculate the mean of foreground and background for each possible threshold;

[0032] Sb3.5, Calculate the inter-class variance: Calculate the inter-class variance for each threshold;

[0033] Sb3.6, Find the best threshold: Find the threshold that maximizes the inter-class variance by traversing all possible thresholds;

[0034] Sb3.7, Generate a binary image: Use the best threshold to segment the image into a binary image.

[0035] Preferably, the S4 step establishes an extraction model to extract image features, and the specific steps are as follows:

[0036] S4.1, Collect data: Collect image data sets containing C and T lines, and label C and T lines in the image to generate training and test sets;

[0037] S4.2, Normalization: Normalize the image pixel values to the range of 0-1;

[0038] S4.3, Build a convolutional neural network: Design a suitable CNN architecture and train the CNN using labeled data to optimize the loss function to improve the accuracy of feature extraction;

[0039] S4.4, Extract C and T line features: Use the trained CNN to extract features of C and T lines;

[0040] S4.5, KNN training: Standardize the features extracted by CNN, and train the KNN model using the processed features and corresponding labels to select an appropriate K value to improve classification accuracy;

[0041] S4.6, Threshold setting: Establish a standard dictionary for C and T lines according to the standard, and adjust the threshold range according to the standard dictionary;

[0042] S4.7, Evaluation and optimization: Use cross-validation method to evaluate the performance of the model to ensure the rationality of the threshold setting, and regularly monitor the performance of the model, collect feedback and make adjustments.

[0043] Preferably, the CNN architecture established in the S4.3 step comprises an input layer, a convolutional layer, an activation layer, a pooling layer and a fully connected layer.

[0044] Preferably, the S5 step uses Matplotlib or Plotly to generate a chart of the detection result, and the detection result is displayed, and the detection chart comprises detection time, sample information and detection result.

[0045] Compared with the prior art, the present application has the following advantages:

[0046] The present application integrates a complete application program, uses a mobile phone or iPad to take a colloidal gold detection card, uploads the image to a cloud server for processing, identifies one or more regions containing the colloidal gold detection card C, T line, reads and calculates the length, width and color value of the C, T line, compares the extracted C, T line value with a standard dictionary, judges the content of the target substance in the sample, records the detection result, and uploads the data to a food safety big data platform, without relying on professional detection instruments, simple operation, humanized use, convenient for food enterprises, farmers' markets, food workshops and families to self-check and screen food, and protect food safety.

[0047] Before identifying the C, T line region in the image and extracting the contour of the C, T line, the OTSU threshold segmentation algorithm is used to segment the image, the best threshold value is obtained, then each pixel in the gray image is compared with the best threshold value, and the image is segmented into foreground and background, so that the noise interference in the image can be reduced before edge detection, the detection algorithm can be more focused on important edge features, the segmented image has clearer edges, the detection accuracy is improved, and the binary image after segmentation makes the contour more obvious, so that the contour extracted by the contour detection algorithm is more complete, and the subsequent feature analysis is facilitated. BRIEF DESCRIPTION OF DRAWINGS

[0048] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor. Among them:

[0049] Figure 1 The method step block diagram of the present application;

[0050] Figure 2 The image preprocessing step block diagram of the present application;

[0051] Figure 3 The region detection step block diagram of the present application;

[0052] Figure 4 This is a block diagram of the steps of using the OTSU threshold segmentation algorithm to segment an image in the present invention;

[0053] Figure 5 This is a flowchart of the steps for extracting image features by establishing an extraction model according to the present invention. DETAILED DESCRIPTION

[0054] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0055] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0056] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0057] As attached Figure 1 To the attached Figure 5 As shown:

[0058] Example 1: This example provides a digital identification method suitable for colloidal gold detection card detection, and the specific steps are as follows:

[0059] S1. Image acquisition: Use a mobile phone or iPad to take a picture of the colloidal gold test card. After ensuring that the image is clear and the light is even, upload the image to the cloud server for processing.

[0060] S2, image preprocessing: use image processing technology to process the image;

[0061] S3, region detection: use edge detection algorithm to identify C and T line regions in the image, and then use contour detection algorithm to extract the contours of C and T lines;

[0062] S4, feature extraction: calculate the length, width and color values ​​of C and T lines, and set the threshold range of CT lines according to the standard dictionary;

[0063] S5. Data comparison and analysis: Compare the extracted C and T line values ​​with the standard dictionary to determine the content of the target substance in the sample, record the test results, and upload the data to the food safety big data platform;

[0064] S6, System integration: integrate S2-S5 into a complete application, provide a user-friendly interface for users to upload images and view detection results.

[0065] Specifically, the specific steps of image preprocessing in S2 are as follows:

[0066] S2.1, Denoising: use a Gaussian filter to smooth the image to reduce noise, the formula is as follows:

[0067] , wherein is a Gaussian function, is the standard deviation, is the pixel coordinate;

[0068] S2.2, Grayscale: convert color images to grayscale images.

[0069] Specifically, the specific steps of region detection in S3 are as follows:

[0070] Sa3.1, Calculate the gradient: use the Sobel operator to calculate the gradient amplitude and direction of the image, the formula of the gradient amplitude is as follows:

[0071] , wherein is the gradient amplitude, indicating the intensity change of the image at a certain point, is the gradient of the image in the horizontal direction, obtained by horizontal edge detection of the image, is the gradient of the image in the vertical direction, obtained by vertical edge detection of the image;

[0072] The formula of the gradient direction is as follows:

[0073] , wherein is the gradient direction, indicating the edge direction of the image at a certain point, is the arctangent function, used to calculate the angle corresponding to the ratio of given and ;

[0074] Sa3.2, Non-maximum suppression: non-maximum suppression is performed on the gradient amplitude, and local maximum values are retained, and non-edge pixels are suppressed;

[0075] Sa3.3, Double threshold processing: set high and low thresholds to divide edges into strong edges, weak edges and non-edges;

[0076] Sa3.4, Edge connection: connect strong edges and weak edges to form complete edges;

[0077] Sa3.5, Extract contours: use the findContours function of OpenCV to extract the contours in the edge image;

[0078] Sa3.6, Draw contours: use the drawContours function to draw the extracted contours on the original image.

[0079] Specifically, in the step of Sa3.2, it is determined whether to retain the pixel by comparing the current pixel with the adjacent pixels in the gradient direction of the pixel.

[0080] Specifically, in the step of Sa3.3, the strong edge is the pixel greater than the high threshold value, the weak edge is the pixel between the high threshold value and the low threshold value, and the non-edge is the pixel less than the low threshold value.

[0081] Specifically, in the step of Sa3.4, if the weak edge is connected with the strong edge, the weak edge is retained, otherwise the weak edge is suppressed.

[0082] Specifically, in the step of S5, Matplotlib or Plotly is used to generate a chart of the detection result to show the detection result, and the detection chart includes detection time, sample information and detection result.

[0083] As can be seen from the above, by integrating a complete application program, after the colloidal gold detection card is photographed by a mobile phone or iPad, the image is uploaded to a cloud server for processing, an edge detection algorithm is used to identify the C and T line regions in the image, and then a contour detection algorithm is used to extract the contours of the C and T lines, so as to identify one or more regions containing the C and T lines of the colloidal gold detection card, calculate the length, width and color value of the C and T lines, compare the extracted C and T line values with a standard dictionary, judge the content of the target substance in the sample, record the detection result, upload the data to a food safety big data platform, and generate a detection chart for personnel to view the detection situation, without relying on professional detection instruments, simple operation, humanized use, convenient for food enterprises, farmers' markets, food workshops and families to conduct self-detection and screening of food, and protection of food safety.

[0084] Embodiment two: this embodiment is basically the same as the previous embodiment, the difference is that, specifically, in the step of S3, before identifying the regions of the C and T lines and extracting the contours of the C and T lines, the OTSU threshold segmentation algorithm is used to segment the image, and the specific steps are as follows:

[0085] Sb3.1, Calculate the image histogram: count the number of pixels of each gray level in the image to generate a gray histogram;

[0086] Sb3.2, Calculate the total number of pixels of the image: calculate the total number of all pixels in the image;

[0087] Sb3.3, Calculate the probability of each gray level: For each gray level, calculate its probability, the formula is as follows:

[0088] wherein is the number of pixels of gray level i, N is the total number of pixels of the image, is the probability of gray level i;

[0089] Sb3.4, Calculate the mean of foreground and background: For each possible threshold, calculate the mean of foreground and background, the formula for calculating the foreground mean is as follows:

[0090] ;

[0091] The formula for calculating the background mean is as follows:

[0092] ;

[0093] Sb3.5, Calculate the inter-class variance: For each threshold, calculate the inter-class variance, the formula is as follows:

[0094] wherein is the inter-class variance, indicating the degree of separation between different classes, t is the threshold currently considered, is the probability of pixels below the threshold t, is the probability of pixels above the threshold t, is the average gray value below the threshold t, is the average gray value above the threshold t;

[0095] Sb3.6, Find the best threshold: By traversing all possible thresholds, find the threshold that maximizes the inter-class variance;

[0096] Sb3.7, Generate a binary image: Use the best threshold to segment the image into a binary image.

[0097] As can be seen from the above, by using the OTSU threshold segmentation algorithm to segment the image before identifying the C and T line regions in the image and extracting the contours of the C and T lines, the optimal threshold is obtained, and then each pixel in the grayscale image is compared with the optimal threshold, if the pixel value is greater than or equal to the threshold, it is set to white, and if the pixel value is less than the threshold, it is set to black, so that the image is segmented into foreground and background, and then the edge detection algorithm is used to identify the C and T line regions in the image, and then the contour detection algorithm is used to extract the contours of the C and T lines, so that the noise interference in the image can be reduced before edge detection, so that the detection algorithm can focus more on important edge features rather than being disturbed by noise, and at the same time, the segmented image will have clearer edges, so that the edge detection algorithm can more accurately identify the edges of the C and T lines, improve the detection accuracy, and the binary image after segmentation will make the contour more obvious, so that the contour detected by the contour detection algorithm is more complete, which is convenient for subsequent feature analysis.

[0098] Embodiment three: this embodiment is basically the same as the previous embodiment, the difference is that the S4 step establishes an extraction model to extract image features, the specific steps are as follows:

[0099] S4.1, collect data: collect image data set containing C and T lines, and label C and T lines in the image to generate training and test set;

[0100] S4.2, normalization processing: normalize the image pixel value to the range of 0-1;

[0101] S4.3, construct convolutional neural network: design a suitable CNN architecture, and train CNN using labeled data, optimize loss function to improve the accuracy of feature extraction;

[0102] S4.4, extract C and T line features: use the trained CNN to extract the features of C and T lines;

[0103] S4.5, KNN training: standardize the features extracted by CNN, and train KNN model using the processed features and corresponding labels to select appropriate K value to improve classification accuracy;

[0104] S4.6, threshold setting: establish the standard dictionary of C and T lines according to the standard, and adjust the threshold range according to the standard dictionary;

[0105] S4.7, evaluation and optimization: use cross-validation method to evaluate the performance of the model to ensure the rationality of the threshold setting, and regularly monitor the performance of the model, collect feedback and make adjustments.

[0106] Specifically, the CNN architecture established in the S4.3 step includes an input layer, a convolution layer, an activation layer, a pooling layer, and a full connection layer.

[0107] As can be seen from the above, the image is input through the input layer, the features are extracted by the convolution layer, the ReLU activation function is used in the activation layer, the feature dimension is reduced by the pooling layer, and then the features of the C and T lines are output through the full connection layer. The CNN is trained using the labeled data, the loss function is optimized to improve the accuracy of feature extraction, and by constructing an identification and analysis model, the combination of the feature extraction capability of the CNN and the classification capability of the KNN can significantly improve the accuracy of calculating the length, width, and color value of the C and T lines. According to the threshold range of the C and T lines set by the standard dictionary, the identification and analysis accuracy and speed are improved.

[0108] The design application is applied to self-checking and screening of food in food enterprises, farmers' markets, food workshops, families, etc. The colloidal gold detection card is photographed by a mobile phone or an iPad and uploaded to a cloud server, one or more regions containing the C and T lines of the colloidal gold detection card are identified, the values of the C and T lines of the colloidal gold detection card are read, and the content of the target in the sample is judged by comparing with a standard dictionary. The detection data can be uploaded to a food safety big data platform for food safety big data analysis. The purpose is to complete the detection and identification by only using a mobile phone and other devices without relying on professional instruments, to reduce the cost, and to facilitate personnel to self-check and screen food.

[0109] Importantly, it should be noted that the configurations and arrangements of the present application shown in a plurality of different exemplary embodiments are merely illustrative. Therefore, the present application is not limited to a particular embodiment, but extends to various modifications that still fall within the scope of the appended claims.

[0110] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not limited. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present application, which should be covered in the scope of the claims of the present application.

Claims

1. A digital recognition method suitable for colloidal gold detection card detection, characterized in that: The specific steps are as follows: S1, image acquisition: use a mobile phone or iPad to take a colloidal gold detection card, ensure that the image is clear and the light is uniform, and then upload the image to a cloud server for processing; S2, image preprocessing: using image processing technology to process the image; The specific steps of image preprocessing in the S2 step are as follows: S2.1, denoising: using a Gaussian filter to smooth the image to reduce noise; Using a Gaussian filter to smooth the image to reduce noise, the formula is as follows: where G(x, y) is a Gaussian function, σ is a standard deviation, and (x, y) is a pixel coordinate; S2.2, grayscale: convert the color image to a grayscale image; S3, region detection: using an edge detection algorithm to identify the C and T line regions in the image, and then using a contour detection algorithm to extract the contours of the C and T lines; The specific steps of region detection in the S3 step are as follows: Sa3.1, gradient calculation: using a Sobel operator to calculate the gradient amplitude and direction of the image; The formula for gradient amplitude is as follows: where G is the gradient magnitude, indicating the degree of intensity change of the image at a point, G x is the gradient of the image in the horizontal direction, obtained by performing horizontal edge detection on the image, G y is the gradient of the image in the vertical direction, obtained by performing vertical edge detection on the image; The formula for gradient direction is as follows: where Θ is the gradient direction, representing the edge direction of the image at a point, tan -1 is the arctangent function, used to calculate the angle corresponding to the ratio of G y and G x . Sa3.2, non-maximum suppression: performing non-maximum suppression on the gradient amplitude, retaining the local maximum value and suppressing non-edge pixels; In the Sa3.2 step, the current pixel is compared with its adjacent pixels in the gradient direction to determine whether to retain the pixel; Sa3.3, double threshold processing: setting a high threshold and a low threshold, and dividing the edge into strong edge, weak edge and non-edge; In the Sa3.3 step, the strong edge is the pixel greater than the high threshold, the weak edge is the pixel between the high threshold and the low threshold, and the non-edge is the pixel less than the low threshold; Sa3.4, edge connection: connecting the strong edge and the weak edge to form a complete edge; In the Sa3.4 step, if the weak edge is connected to the strong edge, the weak edge is retained, otherwise the weak edge is suppressed; Sa3.5, contour extraction: using the findContours function of OpenCV to extract the contours in the edge image; Sa3.6, contour drawing: using the drawContours function to draw the extracted contours on the original image; Before identifying the regions of the C and T lines and extracting the contours of the C and T lines in the S3 step, the image is segmented using the OTSU threshold segmentation algorithm, and the specific steps are as follows: Sb3.1, calculate the image histogram: count the number of pixels in each gray level of the image to generate a gray histogram; Sb3.2, calculate the total number of pixels in the image: calculate the total number of all pixels in the image; Sb3.3, calculate the probability of each gray level: calculate the probability of each gray level; For each gray level, the formula for calculating the probability is as follows: where n(i) is the number of pixels of gray level i, N is the total number of pixels of the image, and P(i) is the probability of gray level i; Sb3.4, calculate the mean of foreground and background: for each possible threshold, calculate the mean of foreground and background; Sb3.5, calculate the inter-class variance: for each threshold, calculate the inter-class variance; For each threshold, the formula for calculating the inter-class variance is as follows: σ 2 (t) = P1(t) - P2(t) - (μ1(t) - μ2(t)) 2 where σ 2 (t) is the inter-class variance, indicating the degree of separation between different classes, t is the threshold currently considered, P1(t) is the probability of pixels below the threshold t, P2(t) is the probability of pixels above the threshold t, μ1(t) is the average gray value below the threshold t, and μ2(t) is the average gray value above the threshold t. Sb3.6, find the best threshold: by traversing all possible thresholds, find the threshold that maximizes the inter-class variance; Sb3.7, generate a binary image: use the best threshold to segment the image into a binary image; The OTSU threshold segmentation algorithm is used to segment the image before identifying the C and T line regions and extracting the contours of the C and T lines. The optimal threshold is obtained by comparing each pixel in the grayscale image with the optimal threshold. If the pixel value is greater than or equal to the threshold, it is set to white; if the pixel value is less than the threshold, it is set to black. The image is then segmented into foreground and background. Subsequently, the edge detection algorithm is used to identify the C and T line regions, and the contour detection algorithm is used to extract the contours of the C and T lines. S4, feature extraction: calculate the length, width and color value of the C and T lines, and set the threshold range of the CT line according to the standard dictionary; The S4 step establishes an extraction model to extract image features, and the specific steps are as follows: S4.1, collect data: collect image data sets containing C and T lines, and label the C and T lines in the image to generate training and test sets; S4.2, normalization processing: normalize the image pixel value to the range of 0-1; S4.3, construct a convolutional neural network: design a suitable CNN architecture and train the CNN using labeled data to optimize the loss function to improve the accuracy of feature extraction; S4.4, extract C and T line features: use the trained CNN to extract the features of the C and T lines; S4.5, KNN training: standardize the features extracted by CNN, and use the processed features and corresponding labels to train the KNN model, and select a suitable K value to improve the classification accuracy; S4.6, threshold setting: establish a standard dictionary of C and T lines according to the standard, and adjust the threshold range according to the standard dictionary; S4.7, evaluation and optimization: use cross-validation method to evaluate the performance of the model to ensure the rationality of the threshold setting, and regularly monitor the performance of the model, collect feedback and make adjustments;S5, data comparison and analysis: compare the extracted C and T line values with the standard dictionary to determine the content of the target substance in the sample, record the detection results, and upload the data to the food safety big data platform; The S5 step uses Matplotlib or Plotly to generate a chart of the detection results, showing the detection results. The detection chart includes detection time, sample information and detection results; S6, system integration: integrate S2-S5 steps into a complete application program, provide a user-friendly interface for users to upload images and view detection results.

2. The digital recognition method suitable for colloidal gold detection card detection according to claim 1, characterized in that: The CNN architecture established in the S4.3 step includes an input layer, a convolutional layer, an activation layer, a pooling layer and a fully connected layer.

Images