Jade category identification method, program product, electronic equipment and storage medium

Abstract

The invention provides a jadeite category identification method, a program product, electronic equipment and a storage medium. The method comprises the following steps: acquiring color data of jadeite to be identified; the color data comprises at least one of color space parameters, tristimulus values and chromaticity coordinates; processing the color data to obtain a color feature vector of the jadeite to be identified; inputting the color feature vector into a trained jadeite classification model to obtain a jadeite category identification result; the jadeite classification model is obtained by training a preset machine learning classification model through sample color feature vectors corresponding to different types of jadeite samples. According to the method, the to-be-identified jadeite is converted into color space parameters, tristimulus values and chromaticity coordinates, so that color observation originally depending on personal experience is converted into digital information, and errors and inconsistent results caused by factors such as experience and light rays in manual identification are reduced. In the whole identification process, only optical measurement needs to be carried out on the surface of the jadeite, destructive operation such as cutting is not needed, and nondestructive testing is achieved.

Description

Technical Field

[0001] This application relates to the field of jadeite identification technology, and more specifically, to a method, program product, electronic device and storage medium for identifying jadeite categories. Background Technology

[0002] Currently, there are two main types of jade identification methods: The first relies on the appraiser's subjective experience in observing characteristics such as color and luster. This method lacks objective standards, resulting in inconsistent results and low efficiency. The second method identifies jade by detecting trace elements. This method usually requires micro- or destructive sampling, making non-destructive identification difficult. Summary of the Invention

[0003] The purpose of this application is to provide a method, program product, electronic device and storage medium for identifying jadeite categories, in order to improve the above-mentioned problems.

[0004] In a first aspect, embodiments of this application provide a method for identifying jadeite categories, comprising: collecting color data of the jadeite to be identified; the color data includes at least one of color space parameters, tristimulus values, and chromaticity coordinates; processing the color data to obtain a color feature vector of the jadeite to be identified; inputting the color feature vector into a trained jadeite classification model to obtain a jadeite category identification result; the jadeite classification model is obtained by training a preset machine learning classification model using the sample color feature vectors corresponding to jadeite samples of different categories.

[0005] In the aforementioned process, firstly, color data of the jadeite is collected, converting it into a set of objective color space parameters, tristimulus values, and chromaticity coordinates. This transforms the previously subjective color observation, which relied on personal experience, into digital information, reducing subjective errors and inconsistencies caused by factors such as experience and lighting in manual identification. Secondly, by standardizing and constructing feature vectors from this color data, a unified and clean data format is provided for subsequent computer analysis, allowing for fair comparison and analysis of data from different batches and instruments. Finally, by inputting the processed color feature vectors into a pre-trained jadeite classification model, rapid identification of the jadeite's origin is achieved. The entire identification process only requires optical measurements of the jadeite surface, eliminating the need for destructive operations such as cutting, thus achieving non-destructive testing. This method reduces reliance on experienced experts and provides a more efficient, consistent, and sample-free jadeite identification solution.

[0006] Optionally, in this embodiment of the application, collecting color data of the jadeite to be identified includes: collecting the spectral reflectance or color signal of the surface of the jadeite to be identified using a color measuring instrument under preset light source and / or preset observation conditions; and converting the spectral reflectance or color signal into at least one of color space parameters, tristimulus values, and chromaticity coordinates based on a colorimetric conversion model.

[0007] In the aforementioned process, color measuring instruments are used to collect the spectral reflectance data of jadeite, reducing subjective observation by the human eye under uncertain ambient light conditions. This makes the source of color information objective and free from human experience or environmental interference. Through a colorimetric conversion model, these raw optical data are precisely calculated into a series of recognized colorimetric parameters, such as color space parameters, tristimulus values, and chromaticity coordinates, achieving the digitization and standardization of color characteristics. Furthermore, the entire data acquisition and conversion process only requires optical irradiation of the jadeite, without any physical contact or damage to the sample, thus achieving non-destructive testing.

[0008] Optionally, in this embodiment, the color space parameters include luminance values, red-green chromaticity values, and yellow-blue chromaticity values ​​in the CIELab color system; the tristimulus values ​​are quantitative indicators used to describe the degree of stimulation of the three primary colors by which the human retina perceives color, and the tristimulus values ​​include red, green, and blue primary colors; the chromaticity coordinates are the chromaticity values ​​in the CIE XYZ color system, where the chromaticity values ​​represent hue and saturation; processing the color data to obtain the color feature vector of the jade to be identified includes: normalizing or standardizing at least one of the color space parameters, tristimulus values, and chromaticity coordinates to obtain preprocessed data; and combining the preprocessed data in a predetermined order to obtain the color feature vector.

[0009] In the above implementation process, the collected raw color parameters are normalized or standardized preprocessed to reduce the inherent imbalance caused by differences in the units and numerical ranges of different color parameters. This allows subsequent machine learning algorithms to consider information in each color dimension equally and fairly. By combining the preprocessed data in a pre-defined fixed order, a color feature vector with a uniform format and clear structure is generated, which meets the standard requirements of machine learning classification models for input data format.

[0010] Optionally, in this embodiment of the application, color data is processed to obtain a color feature vector of the jadeite to be identified, including: calculating color saturation and / or hue angle based on the red-green and yellow-blue values ​​in the color space parameters, and using the color saturation and / or hue angle as derived color features; inputting the color feature vector into a trained jadeite classification model to obtain jadeite category identification results, including: inputting the color feature vector and the derived color features into the trained jadeite classification model to obtain jadeite category identification results; the jadeite classification model is obtained by training a preset machine learning classification model with sample color feature vectors and sample derived color features corresponding to jadeite samples of different categories.

[0011] In the above implementation process, based on the color data, derived color features, such as color saturation and hue angle, are further calculated to directly reflect the vibrancy and hue angle of the color. These derived features provide an additional perspective for describing the color characteristics of jadeite. By combining these derived color features with the original color feature vectors to train the jadeite classification model, the model can refer to more comprehensive and in-depth color information when determining the origin, thus making the model's judgment more comprehensive. When color appearances are similar, the model can capture more subtle differences in color attributes, improving the accuracy of identification.

[0012] Optionally, in this embodiment, before inputting the color feature vector into the trained jade classification model, the method further includes: acquiring jade samples of multiple categories and collecting sample color data of the jade samples; constructing sample color feature vectors of the jade samples based on the sample color data of the jade samples; using the category of the jade sample as a label, and forming a training dataset with the sample color feature vectors corresponding to the jade samples, inputting it into a preset machine learning classification model for training, and then performing model validation on the trained jade classification model to obtain the trained jade classification model; during the training process, cross-validation is used to update the key parameters of the machine learning classification model; the preset machine learning classification model includes a support vector machine model; the key parameters include a penalty factor and kernel function parameters.

[0013] In the above implementation process, these labeled feature vector data are used to train classification models such as support vector machines, and their key parameters are optimized through cross-validation, ultimately resulting in a stable jade classification model. This transforms the core knowledge of jade identification from relying on personal experience into a reusable algorithm model, making the origin identification process more efficient and accurate, and significantly reducing the influence of subjective human factors.

[0014] Optionally, in this embodiment of the application, the trained jade classification model forms a classification decision boundary that distinguishes different jade categories; the classification decision boundary defines the cluster distribution of sample color feature vectors corresponding to different categories of jade samples in the feature space; inputting the color feature vectors into the trained jade classification model to obtain jade category identification results includes: inputting the color feature vectors into the trained jade classification model, the jade classification model determining the category of the jade to be identified based on the classification decision boundary, and generating jade category identification results.

[0015] In the aforementioned implementation process, the trained jadeite classification model can form a clear classification decision boundary. This boundary objectively characterizes the cluster distribution pattern of color features of jadeite from different origins in mathematical space, transforming previously vague empirical judgments into clear geometric classification rules. During actual identification, the color feature vector of the jadeite to be tested is input into the model, which can automatically and accurately determine its category and output the result based on its position relative to this decision boundary. This reduces subjective assumptions and uncertainties in manual identification, resulting in highly repeatable and consistent identification results.

[0016] Optionally, in this embodiment, the categories of jadeite samples include jadeite from a first region and jadeite from a second region; the classification decision boundary is used to divide the first color cluster corresponding to the jadeite from the first region and the second color cluster corresponding to the jadeite from the second region; wherein, in the feature subspace of the color space parameters, the feature vector of the first color cluster has higher statistical values ​​in brightness and red-green hue than that of the second color cluster; or the yellow-blue hue of the first color cluster is lower than that of the second color cluster; in the feature subspace of the tristimulus values, the statistical value of the color feature vector of the first color cluster is higher than that of the second color cluster; in the feature subspace of the chromaticity coordinates, the distribution concentration of the color feature vector of the first color cluster is higher than that of the second color cluster.

[0017] In the aforementioned process, jadeite from different origins exhibits statistical differences in different feature subspaces, such as color space parameters, tristimulus values, and chromaticity coordinates. These differences manifest as variations in brightness and chromaticity values, as well as the concentration and dispersion of chromaticity distribution. These statistical patterns are learned by the model and used to form precise classification decision boundaries. In actual identification, when the color feature vector of the jadeite to be tested is input into the model, the model makes a judgment by determining its association with these "color clusters" defined by statistical patterns. This improves the accuracy and efficiency of identification.

[0018] Optionally, in this embodiment, inputting the color feature vector into a trained jade classification model to obtain jade category identification results includes: collecting surface optical property data of the jade to be identified, including gloss parameters and / or transparency index; obtaining the gloss parameters by measuring the specular reflection intensity of the jade surface using a gloss meter; obtaining the transparency index by measuring the transmitted light of the jade and calculating the intensity ratio of incident light to transmitted light; obtaining the optical feature vector of the jade to be identified based on the surface optical property data; fusing the color feature vector and the optical feature vector to generate an integrated feature; and inputting the integrated feature into the trained jade classification model to obtain jade category identification results; wherein the jade classification model is obtained by training a preset machine learning classification model using sample fusion features corresponding to jade samples of different categories.

[0019] In the above implementation process, the surface gloss and transparency index of jadeite are collected, and these optical property data are used to construct optical feature vectors. By fusing the color feature vector with the optical feature vector, a more comprehensive integrated feature is generated. A classification model trained based on the same integrated feature is then used for identification. This approach not only utilizes color information but also incorporates surface optical properties closely related to the texture and structure of jadeite, providing an additional dimension for origin identification. This improves the overall accuracy and reliability of the identification method.

[0020] Optionally, in this embodiment of the application, the method further includes collecting time-series color data of the jadeite to be identified in chronological order under multiple preset measurement conditions; calculating at least one dynamic stability parameter reflecting the dynamic characteristics of the color of the jadeite to be identified based on the time-series color data; fusing the dynamic stability parameter with the color feature vector to generate a fused feature vector; and inputting the color feature vector into a trained jadeite classification model to obtain the jadeite category identification result, including: inputting the fused feature vector into a trained jadeite classification model to obtain the jadeite category identification result, wherein the jadeite classification model is obtained by training a preset machine learning classification model through the sample fused feature vectors corresponding to jadeite samples of different categories.

[0021] In the aforementioned implementation process, by collecting time-series color data of jadeite, not only is its static color information obtained, but its dynamic behavior of color change with illumination or angle is also captured. Dynamic stability parameters such as color difference and hue fluctuation are calculated from the time-series data, quantifying its dynamic color characteristics. These dynamic parameters are then fused with the original static color feature vector to generate a more comprehensive and dimensionally rich fused feature vector, simultaneously describing the state and changes of jadeite color. In identification, in addition to considering static color, the stability or change pattern of color under different conditions is also taken into account. This helps to identify jadeite from different origins or types that are similar in color under conventional static observation but differ in dynamic optical behavior, further improving the accuracy of the identification method.

[0022] Secondly, this application also provides a jadeite classification identification device, comprising: a data acquisition module for acquiring color data of the jadeite to be identified; the color data including at least one of color space parameters, tristimulus values, and chromaticity coordinates; a feature extraction module for processing the color data to obtain a color feature vector of the jadeite to be identified; and an identification module for inputting the color feature vector into a trained jadeite classification model to obtain a jadeite classification identification result; the jadeite classification model is obtained by training a preset machine learning classification model using the sample color feature vectors corresponding to jadeite samples of different categories. Thirdly, embodiments of this application also provide a computer program product, including computer program instructions, which are executed by a processor to perform the method provided in the first aspect or any implementation thereof.

[0023] Fourthly, embodiments of this application also provide an electronic device, including: a processor and a memory, the memory storing computer program instructions, which are executed by the processor to perform the method provided in the first aspect or any implementation thereof.

[0024] Fifthly, embodiments of this application also provide a computer-readable storage medium storing computer program instructions, which, when executed by a processor, perform the method provided in the first aspect or any implementation thereof.

[0025] This application provides a method, program, electronic device, and storage medium for jadeite classification. First, color data of the jadeite is collected, converting it into a set of objective color space parameters, tristimulus values, and chromaticity coordinates. This transforms subjective color observation, previously reliant on personal experience, into measurable, recordable, and repeatable objective digital information, reducing subjective errors and inconsistencies caused by factors such as experience and lighting in manual identification. Then, through standardized processing and feature vector construction of this color data, a unified and clean data format is provided for subsequent computer analysis, allowing for fair comparison and analysis of data from different batches and instruments. Finally, by inputting the processed color feature vectors into a pre-trained jadeite classification model, rapid identification of the jadeite's origin is achieved. The entire identification process requires only optical measurements of the jadeite surface, eliminating the need for destructive operations such as cutting, thus achieving non-destructive testing. This method reduces reliance on experienced experts and provides a more efficient, consistent, and sample-free jadeite identification solution. Attached Figure Description

[0026] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 A flowchart illustrating a method for identifying jadeite categories provided in this application embodiment; Figure 2 This is a schematic diagram of the structure of the jadeite identification device provided in the embodiments of this application; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0028] The embodiments of the technical solution of this application will now be described in detail with reference to the accompanying drawings. These embodiments are only used to more clearly illustrate the technical solution of this application and are therefore merely examples, and should not be used to limit the scope of protection of this application.

[0029] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit this application.

[0030] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly defined.

[0031] Jadeite is considered an aggregate primarily composed of jadeite mineral, accompanied by various smaller minerals. Its market value is closely related to its type and origin. Although jadeite from Myanmar and Guatemala differs systematically in its mineral composition, coloring elements, and trace element composition, this directly determines the significant differences in their appearance quality and market pricing.

[0032] Jadeite forms in the high-pressure, low-temperature environment of subduction zones, undergoing multiple phases of plate subduction and fluid alteration. This complex process determines its final characteristics and unique properties. Just as a baby's appearance is determined by its mother and DNA, the appearance and essence of jadeite are shaped by its geological environment and formation process. From a genetic perspective, Burmese jadeite forms in a high-pressure, low-temperature metamorphic environment caused by the continuous collision of the Indo-Eurasian plates, where silicate melts directly crystallize into jadeite rock under high temperature and pressure. Guatemalan jadeite, on the other hand, is found in the Motagua Fault Zone formed by the collision of the North American and Caribbean plates, where Cretaceous serpentinite is formed by precipitation or alteration of jadeite and omphacite. These two drastically different mineralization mechanisms result in significant differences in mineral composition, chemical composition, and visual characteristics, with color being particularly prominent: Burmese green jadeite typically has bright, highly saturated colors, primarily in shades of bright green and emerald green, appearing bright and translucent; Guatemalan green jadeite generally has a grayish, darker tone, often with bluish-green hues, resulting in an overall dull appearance. Therefore, color can be one of the key criteria for distinguishing jadeite from these two regions.

[0033] Currently, methods for identifying the origin of jadeite still have significant limitations. Traditional visual identification relies heavily on personal experience, making subjective judgments based on observation of characteristics such as color, luster, and structure. The results are easily influenced by variables such as experience level and ambient lighting, lack a unified and objective evaluation standard, and are difficult to use as valid evidence in commercial disputes.

[0034] With advancements in analytical techniques, laboratories have begun to incorporate scientific instruments to aid in identification. However, current mainstream methods rely excessively on chemical composition analysis, particularly the use of techniques such as laser ablation inductively coupled plasma mass spectrometry (ICP-MS) to determine trace element composition. These methods face two major bottlenecks: first, the high cost of equipment and the highly specialized nature of operation limit their widespread application; second, obtaining accurate compositional data often requires minimally invasive or even destructive sampling, which contradicts the "non-destructive identification" requirement for jadeite as a high-value commodity, significantly restricting its practical application.

[0035] Against this backdrop, there is a need for a new technology for identifying the origin of jadeite that can combine scientific accuracy, non-destructive testing, and high efficiency and practicality, in order to overcome the limitations of existing methodologies, fill industry gaps, and provide reliable technical support for jadeite trade.

[0036] This application provides a method for identifying jadeite categories. First, color data of the jadeite is collected, converting it into a set of objective color space parameters, tristimulus values, and chromaticity coordinates. This transforms subjective color observation, which previously relied on personal experience, into measurable, recordable, and repeatable objective digital information, reducing subjective errors and inconsistencies caused by factors such as experience and lighting in manual identification. Then, by standardizing and constructing feature vectors from this color data, a unified and clean data format is provided for subsequent computer analysis, allowing for fair comparison and analysis of data from different batches and instruments. Finally, by inputting the processed color feature vectors into a pre-trained jadeite classification model, rapid and automatic identification of the jadeite's origin is achieved. The entire identification process only requires optical measurements of the jadeite surface, without any destructive operations such as cutting or scraping, making it completely non-destructive and particularly suitable for identifying high-value jadeite products. This method reduces reliance on experienced experts and provides a more efficient, consistent, and sample-free jadeite identification solution.

[0037] Please see Figure 1 The illustration shows a flowchart of a jadeite identification method provided in an embodiment of this application. The jadeite identification method provided in this embodiment can be applied to electronic devices, which may include physical devices such as servers, PCs, tablets, or smartphones, or virtual devices such as virtual machines or containers. The electronic device can be a single device, a combination of multiple devices, or a cluster of a large number of devices. The jadeite identification method may include: Step S110: Collect color data of the jadeite to be identified; the color data includes at least one of the following: color space parameters, tristimulus values, and chromaticity coordinates.

[0038] Step S120: Process the color data to obtain the color feature vector of the jadeite to be identified.

[0039] Step S130: Input the color feature vector into the trained jade classification model to obtain the jade category identification result; the jade classification model is obtained by training the preset machine learning classification model with the sample color feature vectors corresponding to different categories of jade samples.

[0040] In step S110, the subjective visual perception of the color of jadeite by the human eye is transformed into a series of objective and quantifiable digital parameters using scientific instruments. For example, a high-precision color measurement device, such as a benchtop or portable spectrophotometer, is used, employing a D65 standard light source to simulate sunlight, setting a 10-degree standard observer field of view, and stably placing the jadeite sample on the measurement hole, ensuring that the measurement area is flat, clean, and representative.

[0041] The instrument emits light that shines onto the surface of the jade and collects the spectral information of the reflected light. The instrument's built-in processor or a connected computer calculates and converts this spectral data according to the CIE (International Commission on Illumination) standard colorimetric system. The resulting color data includes color space parameters such as luminance, red-green, and yellow-blue values ​​from the CIELab color table; tristimulus values, which quantify the intensity of the three primary color stimuli perceived by the human retina (red, green, and blue); and chromaticity coordinates in the CIE XYZ color system, representing hue and saturation.

[0042] In one implementation, all these interrelated parameters can be obtained simultaneously through a single measurement, and these color data together constitute a digital fingerprint describing the color attributes of the jadeite sample.

[0043] In step S120, considering that the numerical ranges and dimensions of different parameters in the original color data, such as the brightness value in the color space parameters and the red primary color in the tristimulus values, may vary greatly, directly using them would affect the fairness and accuracy of subsequent models. Therefore, the color data can be preprocessed. The preprocessing method can be standardization, that is, calculating the mean and standard deviation of a certain parameter (such as red primary color) for all samples, and then subtracting the mean from the parameter value of each sample and dividing by the standard deviation, so that the processed parameter data conforms to a standard normal distribution with a mean of 0 and a standard deviation of 1, thereby eliminating the influence of dimensions.

[0044] After preprocessing, all selected color parameters are arranged in a fixed order and combined into a one-dimensional array. For example, the standardized brightness value L, red-green saturation value a, yellow-blue saturation value b, red primary color X, green primary color Y, blue primary color Z, hue x, and saturation y can be concatenated in the above order to form a color feature vector containing 8 values. This color feature vector represents the color characteristics of the jadeite to be tested.

[0045] As one implementation method, to further enhance the representational ability of features, new feature vectors can be derived from the basic parameters. For example, color saturation and hue angle can be calculated based on the red-green value a and the yellow-blue value b, which respectively describe the vividness and hue angle of the color.

[0046] In step S130, the jade classification model is a computer algorithm model that completes the learning process, used to infer the corresponding jade category identification result based on the input color feature vector. It can be understood that the jade category identification result here includes two schemes: one is narrow origin identification, such as the model finally outputting the jade data as Burmese jade, Guatemalan jade, or jade from other regions; the other is broad category identification, such as categories determined based on other attributes of jade, such as different types of jade like glassy jade, icy jade, or dyed jade. These two schemes can be implemented by setting different training labels for the sample data used to train the model. The following content of this application uses the identification of jade from Burma and Guatemala as examples to introduce the scheme.

[0047] Taking origin identification as an example, the training process of a jadeite classification model involves: collecting a large number of jadeite samples with known and accurate origins, such as Burmese jadeite and Guatemalan jadeite; collecting sample color data for each jadeite sample; constructing a sample color feature vector based on the sample color data; and using the origin information as a "label." All labeled sample feature vectors are combined into a training dataset and input into a pre-defined machine learning classification model, such as a Support Vector Machine (SVM), for training. The SVM algorithm automatically finds an optimal classification hyperplane in a high-dimensional feature space, ensuring that the feature vectors of two (or more) classes of samples are separated as clearly and with the largest possible margin.

[0048] During training, techniques such as cross-validation are used to adjust model parameters, prevent overfitting, and ensure the model has good generalization ability. After training, a well-trained jade classification model capable of distinguishing the origin of jade based on color features is obtained.

[0049] In actual identification, simply input the color feature vector obtained from steps S110 and S120 of the jade to be identified into the trained jade classification model. The jade classification model will perform calculations and comparisons according to the classification rules defined within it, and finally output a judgment result, such as "Burmese jade" or "Guatemalan jade", thus completing the identification.

[0050] In the implementation of the above embodiments: First, color data of jadeite is collected, and the jadeite to be identified is converted into a set of objective color space parameters, tristimulus values, and chromaticity coordinates. This transforms the original subjective color observation, which relied on personal experience, into measurable digital information, reducing subjective errors and inconsistencies caused by factors such as experience and lighting in manual identification. Then, through standardized processing and feature vector construction of this color data, a unified and clean data format is provided for subsequent computer analysis, allowing data from different batches and instruments to be compared and analyzed fairly. Finally, by inputting the processed color feature vectors into a pre-trained jadeite classification model, rapid identification of the jadeite's origin is achieved. The entire identification process only requires optical measurements of the jadeite surface, without the need for destructive operations such as cutting, achieving non-destructive testing. This method reduces reliance on experienced identification experts and provides a more efficient, more consistent, and sample-free jadeite identification scheme.

[0051] Optionally, in this embodiment of the application, collecting color data of the jadeite to be identified includes: Under preset light source and / or preset observation conditions, use a color measuring instrument to collect the spectral reflectance or color signal of the jadeite surface to be identified.

[0052] A preset light source refers to a pre-set and standardized artificial lighting source with a known spectral power distribution, such as the CIE D65 standard light source, which simulates average daylight and is an internationally recognized benchmark in the field of color measurement. Preset observation conditions include the standard observer's field of view, defining the spectral range of color perception by the human eye or an instrument simulating the human eye. Color measurement instruments include, for example, spectrophotometers. The data acquisition process involves calibrating the instrument using its built-in standard white board before measurement to ensure the accuracy of the measurement benchmark. After measurement is initiated, the preset light source inside the instrument emits a beam of light containing specific spectral components, illuminating the surface of the jade.

[0053] The instrument's spectral receiver captures the light reflected from the jade's surface and decomposes it into light intensities at different wavelengths, resulting in a spectral reflectance curve, or spectral reflectance. Spectral reflectance represents the ratio of reflected light intensity to incident light intensity at each wavelength. Depending on the instrument configuration, it can directly output a set of color signals preliminarily calculated internally, such as the raw response values ​​of an analog RGB sensor. These raw response values ​​can also serve as the raw data for calculating color data. As an implementation method, to improve data reliability, multiple measurements can be taken from different parts of the same sample, and the average value can be calculated.

[0054] Based on the colorimetric conversion model, spectral reflectance or color signals are converted into at least one of the following: color space parameters, tristimulus values, and chromaticity coordinates. Color space parameters include luminance values, red-green chromaticity values, and yellow-blue chromaticity values ​​in the CIELab color system; tristimulus values ​​are quantitative indicators describing the degree of stimulation of the three primary colors perceived by the human retina, including red, green, and blue primary colors; chromaticity coordinates are chromaticity values ​​in the CIE XYZ color system, representing hue and saturation.

[0055] For example, the measured spectral reflectance data is compared with the CIE-published standard observer color matching function and the relative spectral power distribution data of the standard light source, and a weighted integral is performed within the visible light wavelength range to obtain tristimulus values, namely X, Y, and Z. Among them, the Y value not only represents the green primary color stimulus, but also directly corresponds to the brightness perceived by the human eye.

[0056] Other parameters can be further derived using fixed mathematical formulas: color space parameters refer to the coordinate values ​​in the CIELab (i.e., CIELab color system) uniform color space. Color space parameters convert tristimulus values ​​into L (luminance value), a (red-green chromaticity value), and b (yellow-blue chromaticity value) through nonlinear transformation. Color space parameters make the geometric distance between two points in the space as proportional as possible to the color difference perceived by the human eye.

[0057] Chromaticity coordinates are obtained by normalizing the tristimulus values, and the calculation formulas are x = X / (X + Y + Z) and y = Y / (X + Y + Z). Where X represents the primary red color, Y the primary green color, Z the primary blue color, x represents hue, and y represents saturation. x and y determine the hue and saturation of a color on a two-dimensional chromaticity diagram, eliminating the influence of brightness.

[0058] In the implementation of the above embodiments: A color measuring instrument is used to collect the spectral reflectance data of jadeite, reducing subjective observation by the human eye under uncertain ambient light conditions. This makes the source of color information more objective and free from human experience or environmental interference. Through a colorimetric conversion model, these raw optical data are accurately calculated into colorimetric parameters such as color space parameters, tristimulus values, and chromaticity coordinates, achieving the digitization and standardization of color characteristics. Simultaneously, the entire data acquisition and conversion process only requires optical irradiation of the jadeite, without any physical contact or damage to the sample, achieving non-destructive testing.

[0059] Optionally, in this embodiment of the application, the color data is processed to obtain the color feature vector of the jadeite to be identified, including: At least one of the color space parameters, tristimulus values, and chromaticity coordinates is normalized or standardized as a preprocessing step to obtain preprocessed data. Normalization involves scaling the data proportionally to fit it into a specific interval, such as [0,1]. Standardization involves converting the data into a distribution with a mean of 0 and a standard deviation of 1. For example, the mean (μ) and standard deviation (σ) of a parameter across all training samples can be calculated first, and then the formula (Value - μ) / σ can be applied to the parameter value for each sample. The main purpose of these two preprocessing steps is to eliminate biases caused by differences in units and original numerical ranges between different color parameters.

[0060] The preprocessed data is combined in a predetermined order to obtain color feature vectors. This predetermined order is a fixed arrangement rule determined before model training. For example, the order can be specified as [L, a, b, X, Y, Z, x, y]. The order can be adjusted according to actual circumstances.

[0061] In the implementation of the above embodiments: the collected raw color parameters are normalized or standardized preprocessed to reduce the inherent imbalance caused by the differences in the units and numerical ranges of different color parameters, so that subsequent machine learning algorithms can consider the information in each color dimension equally and fairly. By combining the preprocessed data in a pre-set fixed order, a color feature vector with a uniform format and clear structure is generated, which meets the standard requirements of machine learning classification models for input data format.

[0062] Optionally, in this embodiment of the application, the color data is processed to obtain the color feature vector of the jadeite to be identified, including: Based on the red-green and yellow-blue values ​​in the color space parameters, calculate the color saturation and / or hue angle, and use the color saturation and / or hue angle as derived color features.

[0063] The red-green and yellow-blue chromaticity values ​​in color space parameters are specifically the red-green chromaticity value 'a' and the yellow-blue chromaticity value 'b' measured from the CIE Lab color space. The 'a' value represents the chromaticity in the red (positive) and green (negative) directions, while the 'b' value represents the chromaticity in the yellow (positive) and blue (negative) directions. Color saturation (denoted as C) is calculated as: C = √(a² + b²), which geometrically represents the distance of a color point on the chromaticity plane from the origin. A larger value indicates a more vivid and purer color.

[0064] The hue angle (denoted as h) is calculated as: h = arctan(b / a), and the result is corrected according to the quadrants to represent the angular position of the color on the chromaticity ring. For example, 0° leans towards red, 90° towards yellow, 180° towards green, and 270° towards blue. The calculated C and h values ​​are the "derived color characteristics," which describe the essential properties of color more intuitively from different dimensions (vividness and hue angle), supplementing the information carried by the original L, a, and b values.

[0065] The color feature vector is input into the trained jade classification model to obtain the jade category identification results, including: The color feature vector and derived color features are input into the trained jade classification model to obtain the jade category identification result. The jade classification model is obtained by training the preset machine learning classification model with the sample color feature vector and sample derived color features corresponding to different categories of jade samples.

[0066] For example, during the model building phase, a sample color feature vector is generated for each known category of jade sample, along with corresponding sample derived color features. These two parts are then fused along the feature dimension, for example, by concatenation, to generate a new feature vector with expanded dimensions, which is used to train a pre-defined machine learning classification model. The training process enables the model to learn the complex mapping relationship between the basic color parameters and derived features and the origin category.

[0067] In the model application phase, for each piece of jade to be identified, a complete feature vector containing derived features needs to be generated, and then input into the trained jade classification model. The model will make a comprehensive judgment based on all the knowledge learned from training, including its understanding of color saturation and hue angle patterns, to obtain the jade category identification result, thus utilizing richer color information to improve the accuracy of the identification.

[0068] In the implementation of the above embodiments: based on the color data, derived color features, such as color saturation and hue angle, are further calculated to directly reflect the vibrancy and hue angle of the color. These derived features provide an additional perspective for describing the color characteristics of jadeite. By combining these derived color features with the original color feature vectors to train the jadeite classification model, the model can refer to more comprehensive color information when determining the origin, thus making the model's judgment more comprehensive. When color appearances are similar, the model can capture more subtle differences in color attributes, improving the accuracy of identification.

[0069] Optionally, in this embodiment of the application, before inputting the color feature vector into the trained jade classification model, the method further includes: Obtain jadeite samples from multiple categories and collect their color data. This includes samples of Burmese and Guatemalan jadeite certified by authoritative geological institutions or directly obtained from known mines. Sufficient samples of each category should be collected, covering as many common color and texture variations as possible to ensure the representativeness of the subsequent model. Following standard testing procedures, collect the color data of the jadeite samples. Using a calibrated spectrophotometer, measure the clean, flat surface of each sample to obtain objective color data, including at least one of the following: color space parameters, tristimulus values, and chromaticity coordinates.

[0070] Based on the sample color data of jadeite samples, a sample color feature vector is constructed for each jadeite sample. For example, standardization is performed on each parameter (i.e., subtracting the mean of the parameter from all samples and then dividing by the standard deviation) to eliminate dimensional differences. Then, for each sample, these processed parameter values ​​are arranged in a predefined fixed order to generate the sample color feature vector of the jadeite sample.

[0071] The categories of jade samples are used as labels, and the sample color feature vectors corresponding to the jade samples form a training dataset. This dataset is input into a pre-set machine learning classification model for training. After model validation, the trained jade classification model is obtained. During the training process, cross-validation is used to update the key parameters of the machine learning classification model. The pre-set machine learning classification model includes a support vector machine model. The key parameters include the penalty factor and kernel function parameters.

[0072] Each sample's known origin category (e.g., "Myanmar" or "Guatemala") is used as its label, paired with its corresponding sample color feature vector to form a training dataset. This dataset is then input into a pre-defined machine learning classification model for training. During training, cross-validation methods, such as five-fold cross-validation, are used to update the key parameters of the machine learning classification model: the training data is divided into multiple parts, with one part used as the validation set and the rest as the training set in turn. The average performance of the model under different combinations of penalty factors and kernel function parameters (e.g., the gamma value of a Gaussian kernel) is evaluated using methods such as grid search, thereby selecting the optimal parameter combination and effectively preventing overfitting.

[0073] After training, the model is validated using an independent test set to evaluate its accuracy, recall, and other metrics. The model that meets the performance standards is the trained jade classification model that can be put into use.

[0074] In the implementation of the above embodiments: these labeled feature vector data are used to train classification models such as support vector machines, and their key parameters are optimized through cross-validation, ultimately resulting in a stable jadeite classification model. This transforms the core knowledge of jadeite identification from relying on personal experience into a reusable algorithm model, making the origin identification process more efficient and accurate, and significantly reducing the influence of subjective human factors.

[0075] Optionally, in this embodiment, the trained jade classification model forms a classification decision boundary that distinguishes different jade categories; the classification decision boundary defines the cluster distribution of sample color feature vectors corresponding to different categories of jade samples in the feature space.

[0076] The classification decision boundary is a rule used for discrimination determined by the internal mathematical structure of the model after training. For example, for a Support Vector Machine (SVM) model, the classification decision boundary is an optimal hyperplane or surface found in a high-dimensional feature space. Its goal is to separate the color feature vectors of samples belonging to different categories as much as possible by this boundary, and to maximize the distance or interval from the boundary to the nearest sample points on both sides. During training, the algorithm discovers that the feature vector values ​​of samples from the same origin will cluster in specific regions in the feature space, forming a cluster distribution, while clusters from different origins are separated by the decision boundary.

[0077] The color feature vector is input into the trained jade classification model to obtain the jade category identification results, including: The color feature vector is input into the trained jade classification model. Based on the classification decision boundary, the jade classification model determines the category of the jade to be identified and generates the jade category identification result.

[0078] Once the color feature vector of the jade to be identified is obtained, it is input into a trained jade classification model. The model's internal processing mechanism involves mapping the feature vector to the same feature space used during training and calculating its positional relationship relative to the classification decision boundary already embedded in the model. Based on this positional relationship, a logical judgment is made: if the feature vector falls on one side of the boundary, it is classified as belonging to one category (e.g., Myanmar); if it falls on the other side, it is classified as belonging to another category (e.g., Guatemala). Finally, the model outputs this judgment, generating the jade category identification result. The entire process is completed instantly by the computer, achieving automated judgment from data to conclusion.

[0079] In the implementation of the above embodiments: the trained jadeite classification model can form a clear classification decision boundary. This boundary objectively describes the cluster distribution pattern of color characteristics of jadeite from different origins in mathematical space, transforming the originally vague empirical judgment into clear geometric classification rules. During actual identification, the color feature vector of the jadeite to be tested is input into the model, and the model can automatically and accurately determine its category and output the result based on its position relative to the decision boundary. This reduces subjective assumptions and uncertainties in manual identification, resulting in highly repeatable and consistent identification results.

[0080] Optionally, in this embodiment of the application, the categories of jadeite samples include jadeite from a first region and jadeite from a second region; The classification decision boundary is used to divide the first color cluster corresponding to the jadeite in the first region and the second color cluster corresponding to the jadeite in the second region; wherein, in the feature subspace of the color space parameters, the feature vector of the first color cluster has higher statistical values ​​in brightness and red-green quality than that of the second color cluster; or the yellow-blue quality of the first color cluster is lower than that of the second color cluster.

[0081] The feature subspace of color space parameters refers to the data space formed by extracting dimensions related to the brightness value L, red-green hue value a, and yellow-blue hue value b from the complete color feature vector. Statistical values ​​here typically refer to measures of central tendency such as the mean and median. After model training, the means of L and a can be calculated for all samples in the training set classified as jadeite from the first region (e.g., Burmese jadeite); similarly, the means of the corresponding parameters for jadeite from the second region (e.g., Guatemalan jadeite) can be calculated. By comparison, it can be verified that these two means of the former cluster are usually significantly higher than those of the latter, or the mean b of the former is lower than that of the latter.

[0082] In the feature subspace of the tristimulus values, the statistical value of the color feature vector of the first color cluster is higher than that of the second color cluster. This means that the overall mean values ​​of the X, Y, and Z values ​​of the samples in the first color cluster are higher than those in the second color cluster. This is an objective phenomenon observed through statistical analysis of the tristimulus value data of all training samples during the model training phase. For example, Burmese jade samples, due to their generally brighter and more vibrant colors, have a relatively higher total amount of reflected or transmitted light stimulation, resulting in generally larger calculated X, Y, and Z values. This statistical regularity, as an important separability feature, is captured by machine learning algorithms and used to optimize the position and shape of the classification decision boundary, thereby enhancing the model's ability to make discriminations based on the original light stimulus data.

[0083] In the feature subspace of chromaticity coordinates, the distribution concentration of color feature vectors of the first color cluster is higher than that of the second color cluster.

[0084] Distribution concentration measures the dispersion of data points and can be quantified by calculating the standard deviation or variance of the x and y values ​​of all sample points in a cluster. A smaller standard deviation indicates a more concentrated distribution. During model analysis and validation, it was observed that the first color cluster, representing Burmese jadeite, had a relatively compact distribution of sample points on the xy-plane; while the second color cluster, representing Guatemalan jadeite, had a relatively dispersed distribution of sample points. This difference in distribution concentration reflects the different degrees of inherent consistency in hue and saturation characteristics between the two types of jadeite, providing further evidence for the precise delineation of the classification decision boundary.

[0085] In the implementation of the above embodiments: jadeite from different origins exhibits statistical differences in different feature subspaces, such as color space parameters, tristimulus values, and chromaticity coordinates. These differences manifest as variations in brightness and chromaticity values, as well as the concentration and dispersion of chromaticity distribution. These statistical patterns are learned by the model and used to form precise classification decision boundaries. In actual identification, when the color feature vector of the jadeite to be tested is input into the model, the model makes a judgment by determining its association with these "color clusters" defined by statistical patterns. This improves the accuracy and efficiency of identification.

[0086] Optionally, in this embodiment of the application, the color feature vector is input into a trained jade classification model to obtain jade category identification results, including: The surface optical properties data of the jadeite to be identified are collected, including gloss parameters and / or transparency index. The gloss parameters are obtained by measuring the intensity of specular reflection light on the surface of the jadeite using a gloss meter. The transparency index is obtained by measuring the transmitted light of the jadeite and calculating the intensity ratio of incident light to transmitted light.

[0087] Methods for obtaining gloss parameters: A gloss meter can be used to emit a beam of light onto the polished surface of jadeite at a fixed standard incident angle and receive the specular reflection. The light intensity is measured by a photoelectric sensor. This intensity value is directly output as a quantitative parameter in units of gloss after instrument calibration and calculation. The higher the value, the brighter and more mirror-like the surface. Transparency index measurement requires a dark room environment. A stable light source is used to perpendicularly incident from one side of a jadeite slice or cabochon. A light intensity meter is used on the other side to measure the transmitted light intensity. The index is obtained by calculating the ratio of incident light to transmitted light intensity (i.e., transmittance). The higher the value, the better the translucency of the jadeite.

[0088] Based on surface optical property data, the optical feature vector of the jadeite to be identified is obtained. First, the gloss parameter and transparency index are standardized to bring them into a numerical range similar to the color characteristics. Then, the processed gloss parameter values ​​and transparency index values ​​are arranged and combined in a predefined fixed order, such as [gloss, transparency], to form a one-dimensional array containing several values. This array is the optical feature vector representing the surface optical properties of the jadeite.

[0089] Color feature vectors and optical feature vectors are fused to generate integrated features. Fusion methods include direct concatenation, weighted fusion, or attention-based fusion.

[0090] The integrated features are input into the trained jade classification model to obtain the jade category identification result; the jade classification model is obtained by training the preset machine learning classification model with the sample fusion features corresponding to jade samples of different categories.

[0091] During the learning phase, the model has already mastered the correlation between the combination patterns of color and surface optical properties and the origin. In actual identification, the integrated features of the jadeite to be tested are input into this model, and the model will make a judgment based on the learned comprehensive patterns and obtain the jadeite category identification result.

[0092] In the implementation of the above embodiments: the surface gloss and transparency index of jadeite are collected, and these optical property data are used to construct an optical feature vector. By fusing the color feature vector with the optical feature vector, a more comprehensive integrated feature is generated. Identification is performed using a classification model trained on the same fused feature. This not only utilizes color information but also incorporates surface optical properties closely related to the texture and structure of jadeite, providing an additional dimension for origin identification. This improves the overall accuracy and reliability of the identification method.

[0093] Considering that jadeite from different origins exhibits variations in the types, valence states, and forms of its color-causing ions, this not only leads to static color differences but may also result in dynamic color responses (such as graying or color-changing effects) under different lighting conditions or viewing angles. This is related to the optical effects (such as fluorescence) brought about by the "seed" and "water" of jadeite, and the "seed" and "water" are potentially associated with the origin. This application introduces dynamic colorimetric characteristics as a new identification dimension. By measuring the temporal response of jadeite color under specific controllable change conditions, parameters of its color stability or change pattern are extracted to assist or enhance origin identification.

[0094] Optionally, in this embodiment of the application, the method further includes: Under multiple preset measurement conditions, the temporal color data of the jadeite to be identified were collected in chronological order.

[0095] For example, for the same jadeite sample, multiple color measurements can be performed continuously using the same or synchronously controlled color measuring instruments under at least two preset lighting or observation geometry conditions to obtain a series of time-related color data sequences. Preset measurement conditions include switching between different standard light sources, such as sequentially using a simulated sunlight source, a simulated incandescent lamp source, and an ultraviolet light source to illuminate the same point, or changing the relative relationship between the incident light and the observation angle by rotating the sample stage under a fixed light source.

[0096] The measuring instrument or the computer linked to it will record the spectral reflectance or color signal data collected at each time point under each condition in chronological order. This series of ordered data constitutes the time-series color data.

[0097] Based on time-series color data, at least one dynamic stability parameter reflecting the dynamic characteristics of the color of the jadeite to be identified is calculated.

[0098] For each color signal at different times under each condition, the corresponding color data is calculated according to the colorimetric conversion model. The dynamic stability parameter is calculated by comparing the parameter values ​​at different times or under different conditions. For example, the color dynamic characteristics can be reflected in: calculating the difference between the maximum and minimum hue angle of a sample at different times under the same conditions; or calculating the response rate of brightness changing with time under the same conditions; or calculating the color difference of colors measured under different light sources.

[0099] The dynamic stability parameter is fused with the color feature vector to generate a fused feature vector. Fusion methods include direct concatenation, weighted fusion, or attention-based fusion. Taking direct concatenation as an example, the color feature vector representing static color attributes and the dynamic stability parameter representing dynamic response characteristics are treated as a whole and combined in a predetermined order, for example, static features first and dynamic parameters second, to form a new, higher-dimensional feature array. This new array is the fused feature vector.

[0100] The color feature vector is input into the trained jade classification model to obtain the jade category identification results, including: The fused feature vector is input into the trained jade classification model to obtain the jade category identification result. The jade classification model is obtained by training the preset machine learning classification model with the sample fused feature vectors corresponding to jade samples of different categories.

[0101] The jade classification model used here is specifically trained using sample fusion feature vectors. During the training phase, the input is a sample fusion feature vector composed of the static color feature vectors of jade samples from known origins and their corresponding dynamic stability parameters. The model learns and establishes a mapping relationship between this composite feature and the origin category. Therefore, in actual identification, the fusion feature vector of the jade to be identified is input into this model, and the model can perform comprehensive reasoning based on the learned complex patterns that simultaneously cover static and dynamic color rules, ultimately outputting the jade category identification result.

[0102] In the implementation of the above embodiments: by collecting time-series color data of jadeite, not only is its static color information obtained, but also its dynamic behavior of color change with illumination or angle is captured. Dynamic stability parameters such as color difference and hue fluctuation are calculated from the time-series data, quantifying its dynamic color characteristics. These dynamic parameters are fused with the original static color feature vector to generate a more comprehensive and dimensional fused feature vector, which describes the state and changes of jadeite color. In identification, in addition to considering static color, the stability or change pattern of color under different conditions is also considered. This helps to identify jadeite from different origins or types that are similar in color under conventional static observation but have different dynamic optical behaviors, further improving the accuracy of the identification method.

[0103] Please see Figure 2 The diagram shown is a structural schematic of the jadeite identification device provided in an embodiment of this application; this embodiment of the application provides a jadeite identification device 200, including: The acquisition module 210 is used to acquire color data of the jadeite to be identified; the color data includes at least one of color space parameters, tristimulus values ​​and chromaticity coordinates. The feature extraction module 220 is used to process the color data to obtain the color feature vector of the jadeite to be identified; The identification module 230 is used to input the color feature vector into the trained jade classification model to obtain the jade category identification result; the jade classification model is obtained by training a preset machine learning classification model with the sample color feature vectors corresponding to jade samples of different categories. Optionally, in the embodiments of this application, the jadeite classification identification device 200 and the acquisition module 210 are specifically used to acquire the spectral reflectance or color signal of the surface of the jadeite to be identified using a color measuring instrument under preset light source and / or preset observation conditions; and based on the colorimetric conversion model, convert the spectral reflectance or color signal into at least one of color space parameters, tristimulus values ​​and chromaticity coordinates.

[0104] Optionally, in this embodiment, the jadeite classification device 200 includes color space parameters such as luminance, red-green chromaticity, and yellow-blue chromaticity in the CIELab color system; tristimulus values ​​are quantitative indicators used to describe the degree of three primary color stimuli in human retinal color perception, including red, green, and blue primary colors; chromaticity coordinates are chromaticity values ​​in the CIEXYZ color system, where chromaticity values ​​represent hue and saturation; and a feature extraction module 220 is specifically used to perform normalization or standardization preprocessing on at least one of the color space parameters, tristimulus values, and chromaticity coordinates to obtain preprocessed data; and to combine the preprocessed data in a predetermined order to obtain a color feature vector.

[0105] Optionally, in this embodiment, the jadeite category identification device 200 and the feature extraction module 220 are specifically used to calculate color saturation and / or hue angle based on the red-green and yellow-blue values ​​in the color space parameters, and use the color saturation and / or hue angle as derived color features; input the color feature vector into a trained jadeite classification model to obtain jadeite category identification results, including: inputting the color feature vector and the derived color features into the trained jadeite classification model to obtain jadeite category identification results; the jadeite classification model is obtained by training a preset machine learning classification model with the sample color feature vectors and sample derived color features corresponding to jadeite samples of different categories.

[0106] Optionally, in this embodiment, the jadeite category identification device 200 further includes a model training module, used to acquire jadeite samples of multiple categories and collect sample color data of the jadeite samples; construct sample color feature vectors of the jadeite samples based on the sample color data of the jadeite samples; use the category of the jadeite sample as a label, and form a training dataset with the sample color feature vectors corresponding to the jadeite samples, input it into a preset machine learning classification model for training, and perform model validation on the trained jadeite classification model to obtain a trained jadeite classification model; during the training process, cross-validation is used to update the key parameters of the machine learning classification model; the preset machine learning classification model includes a support vector machine model; the key parameters include a penalty factor and kernel function parameters.

[0107] Optionally, in this embodiment, the jadeite category identification device 200 has a trained jadeite classification model that forms a classification decision boundary to distinguish different jadeite categories; the classification decision boundary defines the cluster distribution of sample color feature vectors corresponding to different categories of jadeite samples in the feature space; the identification module 230 is specifically used to input the color feature vectors into the trained jadeite classification model, and the jadeite classification model determines the category of the jadeite to be identified based on the classification decision boundary and generates a jadeite category identification result.

[0108] Optionally, in this embodiment, the jadeite classification identification device 200 classifies jadeite samples into categories including jadeite from a first region and jadeite from a second region. A classification decision boundary is used to delineate a first color cluster corresponding to jadeite from the first region and a second color cluster corresponding to jadeite from the second region. Specifically, in the feature subspace of color space parameters, the feature vector of the first color cluster has higher statistical values ​​in brightness and red-green hue than that of the second color cluster; or the yellow-blue hue of the first color cluster is lower than that of the second color cluster. In the feature subspace of tristimulus values, the statistical value of the color feature vector of the first color cluster is higher than that of the second color cluster. In the feature subspace of chromaticity coordinates, the distribution concentration of the color feature vector of the first color cluster is higher than that of the second color cluster.

[0109] Optionally, in this embodiment, the jadeite classification device 200 and the classification module 230 are specifically used to collect surface optical characteristic data of the jadeite to be identified. The surface optical characteristic data includes gloss parameters and / or transparency index. The gloss parameters are obtained by measuring the specular reflection intensity of the jadeite surface using a gloss meter. The transparency index is obtained by measuring the transmitted light of the jadeite and calculating the intensity ratio of incident light to transmitted light. Based on the surface optical characteristic data, an optical feature vector of the jadeite to be identified is obtained. The color feature vector and the optical feature vector are fused to generate an integrated feature. The integrated feature is input into a trained jadeite classification model to obtain the jadeite classification result. The jadeite classification model is obtained by training a preset machine learning classification model using the sample fusion features corresponding to jadeite samples of different categories.

[0110] Optionally, in this embodiment, the jadeite category identification device 200 further includes a temporal feature extraction module, used to collect temporal color data of the jadeite to be identified in chronological order under multiple preset measurement conditions; calculate at least one dynamic stability parameter reflecting the dynamic characteristics of the color of the jadeite to be identified based on the temporal color data; fuse the dynamic stability parameter with the color feature vector to generate a fused feature vector; and the identification module 230 is specifically used to input the fused feature vector into a trained jadeite classification model to obtain the jadeite category identification result, wherein the jadeite classification model is obtained by training a preset machine learning classification model through the sample fused feature vectors corresponding to jadeite samples of different categories.

[0111] It should be understood that this device corresponds to the jadeite classification identification method embodiment described above, and is capable of performing the various steps involved in the above method embodiment. The specific functions of this device can be found in the description above, and detailed descriptions are omitted here to avoid repetition. The device includes at least one software functional module that can be stored in memory or embedded in the device's operating system (OS) in the form of software or firmware.

[0112] Please see Figure 3 The diagram shows a structural schematic of an electronic device provided in an embodiment of this application. An electronic device 300 provided in this application includes a processor 310 and a memory 320. The memory 320 stores machine-readable instructions executable by the processor 310. When the machine-readable instructions are executed by the processor 310, the method described above is performed.

[0113] Figure 3 The components shown can be implemented using hardware, software, or a combination thereof. Electronic device 300 may be a physical device, such as a server or PC, or a virtual device, such as a virtual machine or virtualization container. Furthermore, electronic device 300 is not limited to a single device; it can be a combination of multiple devices or a cluster of numerous devices.

[0114] This application also provides a storage medium storing a computer program, which is executed by a processor to perform the above-described method.

[0115] The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0116] This application also provides a computer program product, including computer program instructions, which are executed by a processor to perform the method described above.

[0117] It should be understood that the disclosed apparatus and methods can also be implemented in other ways, given the several embodiments provided in this application. The apparatus embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, or they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0118] In addition, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0119] The above description is only an optional implementation of the embodiments of this application, but the protection scope of the embodiments of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the embodiments of this application should be covered within the protection scope of the embodiments of this application.

Claims

1. A method for identifying jadeite categories, characterized in that, include: Collect color data of the jadeite to be identified; the color data includes at least one of color space parameters, tristimulus values, and chromaticity coordinates. The color data is processed to obtain the color feature vector of the jadeite to be identified; The color feature vector is input into the trained jade classification model to obtain the jade category identification result; the jade classification model is obtained by training a preset machine learning classification model with the sample color feature vectors corresponding to different categories of jade samples.

2. The method according to claim 1, characterized in that, Collect color data of the jadeite to be identified, including: Under preset light source and / or preset observation conditions, the spectral reflectance or color signal of the jadeite surface to be identified is collected using a color measuring instrument; Based on the colorimetric conversion model, the spectral reflectance or color signal is converted into at least one of the color space parameters, tristimulus values, and chromaticity coordinates.

3. The method according to claim 1, characterized in that, The color space parameters include the luminance value, red-green chromaticity value, and yellow-blue chromaticity value in the CIELab color system; the tristimulus values ​​are quantitative indicators used to describe the degree of stimulation of the three primary colors by which the human retina perceives color, and the tristimulus values ​​include the red primary color, the green primary color, and the blue primary color; the chromaticity coordinates are the chromaticity values ​​in the CIE XYZ color system, and the chromaticity values ​​represent hue and saturation; The color data is processed to obtain the color feature vector of the jadeite to be identified, including: Normalize or standardize at least one of the color space parameters, tristimulus values ​​and chromaticity coordinates to obtain preprocessed data; The preprocessed data is combined in a predetermined order to obtain the color feature vector.

4. The method according to claim 3, characterized in that, The color data is processed to obtain the color feature vector of the jadeite to be identified, including: Based on the red-green and yellow-blue values ​​in the color space parameters, calculate the color saturation and / or hue angle, and use the color saturation and / or hue angle as derived color features; The color feature vector is input into the trained jade classification model to obtain the jade category identification result, including: The color feature vector and the derived color feature are input into the trained jade classification model to obtain the jade category identification result; the jade classification model is obtained by training a preset machine learning classification model with the sample color feature vector and sample derived color feature corresponding to different categories of jade samples.

5. The method according to claim 1, characterized in that, Before inputting the color feature vector into the trained jade classification model, the method further includes: Obtain jade samples from multiple categories and collect sample color data of the jade samples; Based on the sample color data of the jade sample, construct the sample color feature vector of the jade sample; The category of the jade sample is used as the label, and the sample color feature vector corresponding to the jade sample constitutes the training dataset. This dataset is input into a preset machine learning classification model for training. After model validation, the trained jade classification model is obtained. During the training process, cross-validation is used to update the key parameters of the machine learning classification model. The preset machine learning classification model includes a support vector machine model. The key parameters include a penalty factor and a kernel function parameter.

6. The method according to claim 1, characterized in that, The trained jade classification model forms a classification decision boundary that distinguishes different jade categories; the classification decision boundary defines the cluster distribution of sample color feature vectors corresponding to different categories of jade samples in the feature space; The color feature vector is input into the trained jade classification model to obtain the jade category identification result, including: The color feature vector is input into the trained jade classification model, which determines the category of the jade to be identified based on the classification decision boundary and generates the jade category identification result.

7. The method according to claim 6, characterized in that, The jadeite samples are categorized into first-region jadeite and second-region jadeite; The classification decision boundary is used to divide the first color cluster corresponding to the jadeite in the first region and the second color cluster corresponding to the jadeite in the second region; wherein, in the feature subspace of the color space parameters, the feature vector of the first color cluster has higher statistical values ​​in brightness and red-green hue than the second color cluster; or the yellow-blue hue of the first color cluster is lower than that of the second color cluster. In the feature subspace of the tristimulus values, the statistical value of the color feature vector of the first color cluster is higher than that of the second color cluster; In the feature subspace of the chromaticity coordinates, the distribution concentration of the color feature vectors of the first color cluster is higher than that of the second color cluster.

8. The method according to claim 1, characterized in that, The color feature vector is input into the trained jade classification model to obtain the jade category identification result, including: The surface optical property data of the jadeite to be identified are collected, including gloss parameters and / or transparency index; the gloss parameters are obtained by measuring the specular reflection intensity of the jadeite surface using a gloss meter; the transparency index is obtained by measuring the transmitted light of the jadeite and calculating the intensity ratio of incident light to transmitted light. Based on the surface optical property data, the optical feature vector of the jadeite to be identified is obtained; The color feature vector and the optical feature vector are fused together to generate an integrated feature; The integrated features are input into a trained jade classification model to obtain jade category identification results; wherein, the jade classification model is obtained by training a preset machine learning classification model with sample fusion features corresponding to jade samples of different categories.

9. The method according to claim 1, characterized in that, The method further includes: Under multiple preset measurement conditions, the temporal color data of the jadeite to be identified is collected in chronological order. Based on the time-series color data, at least one dynamic stability parameter reflecting the dynamic characteristics of the color of the jadeite to be identified is calculated. The dynamic stability parameter is fused with the color feature vector to generate a fused feature vector; The color feature vector is input into the trained jade classification model to obtain the jade category identification result, including: The fused feature vector is input into the trained jade classification model to obtain the jade category identification result. The jade classification model is obtained by training a preset machine learning classification model with the sample fused feature vectors corresponding to jade samples of different categories.

10. A computer program product, characterized in that, It includes computer program instructions that are executed by a processor to perform the method as described in any one of claims 1 to 9.

11. An electronic device, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, perform the method as described in any one of claims 1 to 9.

12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions that, when executed by a processor, perform the method as described in any one of claims 1 to 9.

Images