An artificial intelligence detection system using big data analysis

Abstract

The present invention relates to the field of intelligent detection technology, and discloses an artificial intelligence detection system applying big data analysis, comprising a first acquisition module, comprising a projection lamp, for irradiating an emerald with the projection lamp to obtain a first light transmission image of the emerald; a vibration module, for identifying the first light transmission image to obtain a crack width and a crack depth of the emerald; after determining a vibration frequency and a vibration amplitude of the emerald according to the crack width and the crack depth of the emerald, the emerald is vibrated; a second acquisition module, for irradiating the vibrated emerald with the projection lamp to obtain a second light transmission image of the emerald; and a model detection module, for obtaining a characteristic difference matrix of the first light transmission image and the second light transmission image, and inputting the difference matrix into a preset deviation detection model, the deviation detection model outputs a deviation coefficient of the emerald, and the quality grade of the emerald is evaluated according to the deviation coefficient.

Description

Technical Field

[0001] The present invention relates to the field of intelligent detection technology, and more specifically, to an artificial intelligence detection system using big data analysis. Background Art

[0002] Emeralds can be divided into two types according to their sources: natural and synthetic. Natural emeralds are formed under high temperature and high pressure deep in the earth's crust. There are cracks filled with organic liquids and bubbles in the organic liquids. Due to the high temperature and high pressure formation conditions, the bubbles in natural emeralds are usually very stable and will hardly move under external vibrations. In contrast, synthetic emeralds are produced by industrial means to imitate the mineralization process in the natural environment. Although the production process simulates natural conditions as much as possible, due to technical and process limitations, the stability of bubbles inside synthetic emeralds is often not as good as that of natural products, and they are easy to move under vibration. There is currently a physical detection method for the quality assessment of synthetic emeralds. By applying vibration to the sample, experienced appraisers observe and record the relative position changes of bubbles before and after the vibration through a microscope. Based on the degree of bubble movement, the appraiser will evaluate the quality grade of the emerald. However, this traditional manual observation method has the disadvantages of strong subjectivity, low efficiency and difficulty in ensuring consistency. Summary of the invention

[0003] The present invention provides an artificial intelligence detection system using big data analysis, which solves the technical problems raised in the background technology.

[0004] The present invention provides an artificial intelligence detection system using big data analysis, comprising:

[0005] The first acquisition module includes a projection lamp, which is used to illuminate the emerald through the projection lamp to obtain a first light transmission image of the emerald;

[0006] A vibration module is used to identify the first light transmission image to obtain the crack width and crack depth of the emerald; after determining the vibration frequency and vibration amplitude of the emerald according to the crack width and crack depth of the emerald, the emerald is vibrated;

[0007] The second acquisition module is used to illuminate the vibrated emerald through a projection lamp to obtain a second light transmission image of the emerald;

[0008] The model detection module is used to obtain the characteristic difference matrix of the first light transmittance image and the second light transmittance image, and input the characteristic difference matrix into a preset deviation detection model. The deviation detection model outputs the deviation coefficient of the emerald and evaluates the quality grade of the emerald according to the deviation coefficient.

[0009] Furthermore, the first light transmission image is identified to obtain the crack width and crack depth of the emerald, including:

[0010] Step 1, performing edge detection on the first light transmission image to obtain a number of characteristic edges;

[0011] Step 2: Manually mark the feature edges, including:

[0012] Manually matching and marking the feature edge with the outer boundary of the emerald to obtain a first marking;

[0013] Perform shape recognition on the remaining feature edges and perform a second annotation on the circular feature edges;

[0014] Perform a third annotation on the feature edge remaining after the first and second annotations;

[0015] Step 3: determine the maximum number of pixels and the minimum number of pixels between any two feature edges from the third marked feature edges, and sum the pixels of the maximum number of pixels and the minimum number of pixels As the width of the cracks in emeralds;

[0016] Step 4: determine the feature edge with the largest number of pixels and the feature edge with the smallest number of pixels from the third marked feature edge, and sum the pixels of the feature edge with the largest number of pixels and the feature edge with the smallest number of pixels. As the depth of the cracks in the emerald.

[0017] Furthermore, the vibration frequency and amplitude of the emerald are determined according to the crack width and crack depth of the emerald, including:

[0018] Vibration frequency of emerald The calculation formula is as follows:

[0019] ;

[0020] in, represents the elastic modulus of emerald, represents the crack width, Indicates the crack depth, Represents the total number of pixels of the feature edge of the third annotation, Indicates the number of pixels in the area surrounding the edge of the third annotation feature, , , and Represent different linear empirical coefficients, represents the reverse correction coefficient, , Represents the natural base;

[0021] Vibration amplitude of emerald The calculation formula is as follows:

[0022] ;

[0023] in, and Represent different exponential empirical coefficients, , and Represent different linear empirical coefficients, Indicates the maximum vibration amplitude.

[0024] Furthermore, obtaining a characteristic difference matrix of the first light transmittance image and the second light transmittance image includes:

[0025] Step 5, obtaining the pixel value of each pixel of the first light transmittance map and the second light transmittance map, and generating the first matrix and the second matrix respectively according to the pixel value, and performing element-by-element difference calculation on the first matrix and the second matrix to obtain the characteristic difference matrix of the first light transmittance map and the second light transmittance map; wherein, the elements greater than 0 in the characteristic difference matrix are replaced by 1, and the elements less than or equal to 0 are replaced by 0.

[0026] Furthermore, the deviation detection model is as follows:

[0027] Step 6, repeat steps 1-5 to obtain The characteristic difference matrix and manual evaluation deviation coefficient of emeralds;

[0028] Step 7: feature difference matrices as training samples for the deviation detection model. The manual evaluation deviation coefficient is used as the sample label corresponding to the training sample to obtain Set training data;

[0029] Step 8, pass The pre-built initial detection model is trained with the group training data to obtain a deviation detection model; the deviation detection model is reversely updated through the mean error variance loss function; wherein the deviation detection model includes a correction layer, a hidden layer and a mapping layer.

[0030] Further, the correction layer is as follows:

[0031] Step 9, clustering the emeralds according to the crack widths and crack depths of the emeralds to obtain several types of characteristic emeralds;

[0032] Step 10, analyzing each type of characteristic emerald, including:

[0033] Step 101, one emerald of each characteristic type is selected, and magnetic particles are placed in the emerald through micro-openings; wherein the size of the magnetic particles is equal to the average size of the bubbles;

[0034] Step 102, using magnetic force at fixed time intervals Control the movement of magnetic particles within the emerald to a fixed distance and at a fixed time interval Get the light transmission map corresponding to the emerald stone and get the light transmission map collection , and calculate the light perspective map set The difference matrix between adjacent light transmission images in ; wherein the elements greater than 0 in the difference matrix are replaced by 1, and the elements less than or equal to 0 are replaced by 0;

[0035] Step 103, matching the position coordinates of the magnetic particles in the light transmission image at each moment with the corresponding difference matrix, counting the number of elements 1 in the difference matrix, and assigning the number of elements 1 as weights to the corresponding position coordinates;

[0036] Step 11, according to the weight of each position coordinate in the light transmission map of each type of emerald, the characteristic difference matrix is ​​corrected as follows:

[0037] Step 111, performing eight-connectivity analysis on element 1 in the characteristic difference matrix to obtain connected clusters in the characteristic difference matrix, retaining a connected cluster with the largest number of element 1s in the connected clusters, and replacing the remaining connected clusters with element 0 to obtain a replacement matrix;

[0038] Step 112, obtaining the position coordinates of element 1 in the replacement matrix, and calculating the weighted mean and center position coordinates of the corresponding position coordinates; wherein the weighted mean represents the sum of the weights of all elements 1 divided by the number of elements 1, and the center position coordinate represents the center of all elements 1;

[0039] Step 113, dividing the number of elements 1 by the weight mean to obtain the number of updated elements 1; taking the center position coordinate as the center of the circle, calculating the Euclidean distance between the remaining elements 1 and the center position coordinate, and sorting the elements 1 from large to small according to the Euclidean distance to obtain a feature sorting;

[0040] Step 114, replace the corresponding elements 1 in the feature sorting from front to back with 0 until the number of remaining elements 1 is equal to the number of updated elements 1, and then obtain a correction matrix.

[0041] Furthermore, the hidden layer includes convolution units and prediction units, as follows:

[0042] The convolution unit is constructed based on a convolutional neural network. The convolution unit is used to input a correction matrix and output a feature map of the correction matrix.

[0043] The prediction unit is built based on the support vector machine. The feature map is expanded into a feature vector by row. The prediction unit is used to input the feature vector and output the hidden state of the feature vector.

[0044] Furthermore, the mapping layer includes a classifier, as follows:

[0045] The classification space of the classifier represents the deviation coefficient of the emerald; the classifier is used to input the hidden state and map the hidden state to the deviation coefficient of the emerald.

[0046] Furthermore, the quality grade of emerald is evaluated based on the coefficient of deviation, including:

[0047] The deviation coefficient ranges from 0 to 100, and 0-10, 10-20, 20-30, 30-40, 40-50, 50-60, 60-70, 70-80, 80-90 and 90-100 are mapped to 10 quality levels respectively;

[0048] Based on the deviation coefficient of the emerald, the corresponding quality grade of the emerald is determined.

[0049] The beneficial effect of the present invention is that through the systematic analysis of the light transmission image and the characteristics before and after vibration of emeralds, efficient, accurate and standardized quality assessment is achieved. This not only solves the problems of strong subjectivity and low efficiency in traditional detection methods, but also improves the detection accuracy by building a deviation detection model. At the same time, it can grade and evaluate the quality of emeralds, which is suitable for industrial large-scale detection scenarios and significantly improves the intelligence level and consistency of detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a module diagram of an artificial intelligence detection system using big data analysis of the present invention;

[0051] Figure 2 It is a schematic diagram of the projection lamp irradiating emerald in the present invention. DETAILED DESCRIPTION

[0052] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.

[0053] like Figure 1-2 As shown, an artificial intelligence detection system using big data analysis includes:

[0054] The first acquisition module includes a projection lamp, which is used to illuminate the emerald through the projection lamp to obtain a first light transmission image of the emerald;

[0055] A vibration module is used to identify the first light transmission image to obtain the crack width and crack depth of the emerald; after determining the vibration frequency and vibration amplitude of the emerald according to the crack width and crack depth of the emerald, the emerald is vibrated;

[0056] The second acquisition module is used to illuminate the vibrated emerald through a projection lamp to obtain a second light transmission image of the emerald;

[0057] The model detection module is used to obtain the characteristic difference matrix of the first light transmittance image and the second light transmittance image, and input the characteristic difference matrix into a preset deviation detection model. The deviation detection model outputs the deviation coefficient of the emerald and evaluates the quality grade of the emerald according to the deviation coefficient.

[0058] Specifically, the light transmission images are obtained by projecting a light onto the emerald and mapping it onto the display screen. The emerald obtains the first light transmission image (without vibration) and the second light transmission image (after vibration). If the bubbles in the emerald are recognizable, it means that the air pressure inside the emerald is unstable and it is an inferior product.

[0059] In one embodiment of the present invention, identifying the first light transmission image to obtain the crack width and crack depth of the emerald includes:

[0060] Step 1, performing edge detection on the first light transmission image to obtain a number of characteristic edges;

[0061] Step 2: Manually mark the feature edges, including:

[0062] Perform a third annotation on the feature edge remaining after the first and second annotations;

[0063] Step 3: determine the maximum number of pixels and the minimum number of pixels between any two feature edges from the third marked feature edges, and sum the pixels of the maximum number of pixels and the minimum number of pixels As the width of the cracks in emeralds;

[0064] Step 4: determine the feature edge with the largest number of pixels and the feature edge with the smallest number of pixels from the third marked feature edge, and sum the pixels of the feature edge with the largest number of pixels and the feature edge with the smallest number of pixels. As the depth of the cracks in the emerald.

[0065] Specifically, the edge detection algorithm can be a Canny edge detection algorithm. By detecting the edges of the first light transmission image, many characteristic edges are obtained. Usually, there are contours on the surface of the emerald that are thought to be obtained by cutting, so these contours will also form characteristic edges, but because this is artificially processed, through a single match between the artificial characteristic edges and the boundary contours, these characteristic edges can be automatically matched to the boundary contours in all subsequent first light transmission images. When the matching is completed, the remaining characteristic contours also include the characteristic edges of the cracks and the characteristic edges of the bubbles. Since the characteristic edges of the bubbles are circular (approximately circular), there is a significant difference from the characteristic edges of the cracks. Therefore, through shape detection, the characteristic edges of the bubbles can be quickly excluded. The remaining characteristic edges (third annotations) are automatically classified as characteristic edges of the cracks.

[0066] In one embodiment of the present invention, the vibration frequency and vibration amplitude of the emerald are determined according to the crack width and crack depth of the emerald, including:

[0067] Vibration frequency of emerald The calculation formula is as follows:

[0068] ;

[0069] in, represents the elastic modulus of emerald, represents the crack width, Indicates the crack depth, Represents the total number of pixels of the feature edge of the third annotation, Indicates the number of pixels in the area surrounding the edge of the third annotation feature, , , and Represent different linear empirical coefficients, represents the reverse correction coefficient, , Represents the natural base;

[0070] Vibration amplitude of emerald The calculation formula is as follows:

[0071] ;

[0072] in, and Represent different exponential empirical coefficients, , and Represent different linear empirical coefficients, Indicates the maximum vibration amplitude.

[0073] Specifically, since the cracks inside emeralds vary in size (artificially simulating natural mineralization), the vibration frequency and amplitude should be smaller for emeralds with larger internal cracks, and larger for emeralds with smaller internal cracks. This is because the smaller the cracks in emeralds, the better their stability.

[0074] In one embodiment of the present invention, obtaining a characteristic difference matrix of a first light transmittance image and a second light transmittance image includes:

[0075] Step 5, obtaining the pixel value of each pixel of the first light transmittance map and the second light transmittance map, and generating the first matrix and the second matrix respectively according to the pixel value, and performing element-by-element difference calculation on the first matrix and the second matrix to obtain the characteristic difference matrix of the first light transmittance map and the second light transmittance map; wherein, the elements greater than 0 in the characteristic difference matrix are replaced by 1, and the elements less than or equal to 0 are replaced by 0.

[0076] It should be noted that the calculation of the difference matrix is ​​a prior art and will not be described in detail.

[0077] In one embodiment of the present invention, the deviation detection model is as follows:

[0078] Step 6, repeat steps 1-5 to obtain The characteristic difference matrix and manual evaluation deviation coefficient of emeralds;

[0079] Step 7: feature difference matrices as training samples for the deviation detection model. The manual evaluation deviation coefficient is used as the sample label corresponding to the training sample to obtain Set training data;

[0080] Step 8, pass The pre-built initial detection model is trained with the group training data to obtain a deviation detection model; the deviation detection model is reversely updated through the mean error variance loss function; wherein the deviation detection model includes a correction layer, a hidden layer and a mapping layer.

[0081] Specifically, the deviation detection model is a supervised learning model. By learning the inspector's evaluation, the deviation detection model can output an evaluation of the deviation coefficient similar to that of the inspector.

[0082] In one embodiment of the present invention, the correction layer is as follows:

[0083] Step 9, clustering the emeralds according to the crack widths and crack depths of the emeralds to obtain several types of characteristic emeralds;

[0084] Step 10, analyzing each type of characteristic emerald, including:

[0085] Step 101, one emerald of each characteristic type is selected, and magnetic particles are placed in the emerald through micro-openings; wherein the size of the magnetic particles is equal to the average size of the bubbles;

[0086] Step 102, using magnetic force at fixed time intervals Control the movement of magnetic particles within the emerald to a fixed distance and at a fixed time interval Get the light transmission map corresponding to the emerald stone and get the light transmission map collection , and calculate the light perspective map set The difference matrix between adjacent light transmission images in ; wherein the elements greater than 0 in the difference matrix are replaced by 1, and the elements less than or equal to 0 are replaced by 0;

[0087] Step 103, matching the position coordinates of the magnetic particles in the light transmission image at each moment with the corresponding difference matrix, counting the number of elements 1 in the difference matrix, and assigning the number of elements 1 as weights to the corresponding position coordinates;

[0088] Step 11, according to the weight of each position coordinate in the light transmission map of each type of emerald, the characteristic difference matrix is ​​corrected as follows:

[0089] Step 111, performing eight-connectivity analysis on element 1 in the characteristic difference matrix to obtain connected clusters in the characteristic difference matrix, retaining a connected cluster with the largest number of element 1s in the connected clusters, and replacing the remaining connected clusters with element 0 to obtain a replacement matrix;

[0090] Step 112, obtaining the position coordinates of element 1 in the replacement matrix, and calculating the weighted mean and center position coordinates of the corresponding position coordinates; wherein the weighted mean represents the sum of the weights of all elements 1 divided by the number of elements 1, and the center position coordinate represents the center of all elements 1;

[0091] Step 113, dividing the number of elements 1 by the weight mean to obtain the number of updated elements 1; taking the center position coordinate as the center of the circle, calculating the Euclidean distance between the remaining elements 1 and the center position coordinate, and sorting the elements 1 from large to small according to the Euclidean distance to obtain a feature sorting;

[0092] Step 114, replace the corresponding elements 1 in the feature sorting from front to back with 0 until the number of remaining elements 1 is equal to the number of updated elements 1, and then obtain a correction matrix.

[0093] Specifically, the projector illuminates the emerald, and then a light transmission image is obtained on the display screen. Since the light transmission image is formed by the refraction of light inside the emerald, there is a clear difference between the intersection of the refraction of light and the intersection of non-refraction of light. For example, a bubble moves to and a bubble moves to Assume that is a light intersection point, then for The influence range of the light penetration map is much smaller than In the influence range of the light transmission map. Therefore, by conducting a magnetic force experiment on each position of the light transmission map (such as step 10), the weight of each position is obtained, and the light transmission map is corrected by the weight, thereby eliminating the influence of light refraction and obtaining a correction matrix. The correction matrix is ​​more conducive to the recognition of the hidden layer and increases the accuracy of the recognition of the hidden layer.

[0094] In one embodiment of the present invention, the hidden layer includes a convolution unit and a prediction unit, as follows:

[0095] The convolution unit is constructed based on a convolutional neural network. The convolution unit is used to input a correction matrix and output a feature map of the correction matrix.

[0096] The prediction unit is built based on the support vector machine. The feature map is expanded into a feature vector by row. The prediction unit is used to input the feature vector and output the hidden state of the feature vector.

[0097] Specifically, convolutional neural networks and support vector machines are both existing technologies, so they are not described in detail.

[0098] In one embodiment of the present invention, the mapping layer includes a classifier, as follows:

[0099] The classification space of the classifier represents the deviation coefficient of the emerald; the classifier is used to input the hidden state and map the hidden state to the deviation coefficient of the emerald.

[0100] Specifically, the classifier is an existing technology, so it will not be described in detail.

[0101] In one embodiment of the present invention, the quality grade of emerald is evaluated according to the deviation coefficient, including:

[0102] The deviation coefficient ranges from 0 to 100, and 0-10, 10-20, 20-30, 30-40, 40-50, 50-60, 60-70, 70-80, 80-90 and 90-100 are mapped to 10 quality levels respectively;

[0103] Based on the deviation coefficient of the emerald, the corresponding quality grade of the emerald is determined.

[0104] For example, if the deviation coefficient is 68, which is in the range of 60-70, then the corresponding quality grade of the emerald is 7, which is of medium to low quality. The higher the quality grade of the emerald, the worse the quality.

[0105] The above describes an embodiment of the present embodiment, but the present embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present embodiment, ordinary technicians in this field can also make many forms, all of which are within the protection of the present embodiment.

Claims

1. An artificial intelligence detection system using big data analysis, characterized in that: include: The first acquisition module includes a projection lamp, which is used to illuminate the emerald through the projection lamp to obtain a first light transmission image of the emerald; The vibration module is used to identify the first light transmission image to obtain the crack width and crack depth of the emerald, specifically including: Step 1, performing edge detection on the first light transmission image to obtain a number of characteristic edges; Step 2: Manually mark the feature edges, including: Manually matching and marking the feature edge with the outer boundary of the emerald to obtain a first marking; Perform shape recognition on the remaining feature edges and perform a second annotation on the circular feature edges; Perform a third annotation on the feature edge remaining after the first and second annotations; Step 3: determine the maximum number of pixels and the minimum number of pixels between any two feature edges from the third marked feature edges, and sum the pixels of the maximum number of pixels and the minimum number of pixels As the width of the cracks in emeralds; Step 4: determine the feature edge with the largest number of pixels and the feature edge with the smallest number of pixels from the third marked feature edge, and sum the pixels of the feature edge with the largest number of pixels and the feature edge with the smallest number of pixels. As the depth of the cracks in the emerald; After determining the vibration frequency and amplitude of the emerald according to the crack width and crack depth of the emerald, the emerald is vibrated, specifically including: The calculation formula for the vibration frequency f of emerald is as follows: Among them, E represents the elastic modulus of emerald, w represents the crack width, d represents the crack depth, L e Indicates the total number of pixels of the feature edge of the third annotation, S e represents the number of pixels in the area surrounded by the feature edge of the third annotation, α1, α2, α3 and α4 represent different linear empirical coefficients, δ represents the reverse correction coefficient, exp(·)=e(·), and e represents the natural base; The calculation formula for the vibration amplitude A of emerald is as follows: Among them, n and r represent different exponential empirical coefficients, α5, α6 and α7 represent different linear empirical coefficients, and u represents the maximum vibration amplitude; The second acquisition module is used to illuminate the vibrated emerald through a projection lamp to obtain a second light transmission image of the emerald; The model detection module is used to obtain the characteristic difference matrix of the first light transmittance image and the second light transmittance image, and input the characteristic difference matrix into a preset deviation detection model. The deviation detection model outputs the deviation coefficient of the emerald and evaluates the quality grade of the emerald according to the deviation coefficient.

2. The artificial intelligence detection system using big data analysis according to claim 1, characterized in that: Obtaining a characteristic difference matrix of a first light transmission image and a second light transmission image, comprising: Step 5, obtaining the pixel value of each pixel of the first light transmittance map and the second light transmittance map, and generating the first matrix and the second matrix respectively according to the pixel value, and performing element-by-element difference calculation on the first matrix and the second matrix to obtain the characteristic difference matrix of the first light transmittance map and the second light transmittance map; wherein, the elements greater than 0 in the characteristic difference matrix are replaced by 1, and the elements less than or equal to 0 are replaced by 0.

3. The artificial intelligence detection system using big data analysis according to claim 2 is characterized in that: The deviation detection model is as follows: Step 6, repeat steps 1-5 to obtain the characteristic difference matrix and manual evaluation deviation coefficient of K emeralds; Step 7, using K feature difference matrices as training samples of the deviation detection model, and using K manually evaluated deviation coefficients as sample labels corresponding to the training samples, to obtain K groups of training data; Step 8, training the pre-built initial detection model through K groups of training data to obtain a deviation detection model; the deviation detection model is reversely updated through the mean square error loss function; wherein the deviation detection model includes a correction layer, a hidden layer and a mapping layer.

4. The artificial intelligence detection system using big data analysis according to claim 3 is characterized in that: Correction layer, as follows: Step 9, clustering the emeralds according to the crack widths and crack depths of the emeralds to obtain several types of characteristic emeralds; Step 10, analyzing each type of characteristic emerald, including: Step 101, one emerald of each characteristic type is selected, and magnetic particles are placed in the emerald through micro-openings; wherein the size of the magnetic particles is equal to the average size of the bubbles; Step 102, controlling the magnetic particles to move a fixed distance inside the emerald at a fixed time interval R by magnetic force, and obtaining the light transmission image corresponding to the emerald at the fixed time interval R, to obtain a light transmission image set S n , and calculate the light perspective map set S n The difference matrix between adjacent light transmission images in ; wherein the elements greater than 0 in the difference matrix are replaced by 1, and the elements less than or equal to 0 are replaced by 0; Step 103, matching the position coordinates of the magnetic particles in the light transmission image at each moment with the corresponding difference matrix, counting the number of elements 1 in the difference matrix, and assigning the number of elements 1 as weights to the corresponding position coordinates; Step 11, according to the weight of each position coordinate in the light transmission map of each type of emerald, the characteristic difference matrix is ​​corrected as follows: Step 111, performing eight-connectivity analysis on element 1 in the characteristic difference matrix to obtain connected clusters in the characteristic difference matrix, retaining a connected cluster with the largest number of element 1s in the connected clusters, and replacing the remaining connected clusters with element 0 to obtain a replacement matrix; Step 112, obtaining the position coordinates of element 1 in the replacement matrix, and calculating the weighted mean and center position coordinates of the corresponding position coordinates; wherein the weighted mean represents the sum of the weights of all elements 1 divided by the number of elements 1, and the center position coordinate represents the center of all elements 1; Step 113, dividing the number of elements 1 by the weight mean to obtain the number of updated elements 1; taking the center position coordinate as the center of the circle, calculating the Euclidean distance between the remaining elements 1 and the center position coordinate, and sorting the elements 1 from large to small according to the Euclidean distance to obtain a feature sorting; Step 114, replace the corresponding elements 1 in the feature sorting from front to back with 0 until the number of remaining elements 1 is equal to the number of updated elements 1, and then obtain a correction matrix.

5. The artificial intelligence detection system using big data analysis according to claim 4 is characterized in that: The hidden layer includes convolution units and prediction units, as follows: The convolution unit is constructed based on a convolutional neural network. The convolution unit is used to input a correction matrix and output a feature map of the correction matrix. The prediction unit is built based on the support vector machine. The feature map is expanded into a feature vector by row. The prediction unit is used to input the feature vector and output the hidden state of the feature vector.

6. The artificial intelligence detection system using big data analysis according to claim 5, characterized in that: The mapping layer includes classifiers as follows: The classification space of the classifier represents the deviation coefficient of the emerald; the classifier is used to input the hidden state and map the hidden state to the deviation coefficient of the emerald.

7. The artificial intelligence detection system using big data analysis according to claim 1, characterized in that: The quality grade of emeralds is assessed based on the coefficient of variation, including: The deviation coefficient ranges from 0 to 100, and 0-10, 10-20, 20-30, 30-40, 40-50, 50-60, 60-70, 70-80, 80-90 and 90-100 are mapped to 10 quality levels respectively; Based on the deviation coefficient of the emerald, the corresponding quality grade of the emerald is determined.

Images