Search

CN-122023329-A - Gold industry e-commerce SaaS cloud service realization method and system

CN122023329ACN 122023329 ACN122023329 ACN 122023329ACN-122023329-A

Abstract

The invention provides a gold industry electronic commerce SaaS cloud service implementation method and system, which comprise the steps of collecting and extracting hyperspectral feature images, polarized feature images and dimension features of defect mask images of gold products in real time to splice and form dimension-reducing fusion feature vectors, constructing feature manifolds reflecting gold coating oxidation evolution continuity based on distribution situations of the dimension-reducing fusion feature vectors in a low-dimension feature space, selecting and calculating local density entropy and feature vector field divergence of gold coating sample points along an oxidation main axis, and solving the problems that the traditional gold industry electronic commerce SaaS cloud service implementation method is difficult to analyze space-time differences of gold surface oxide layer thickness distribution and coating reflectivity, so that consumers cannot truly observe colors and textures of gold products through images, and development of gold industry electronic commerce is severely restricted.

Inventors

  • LIU HAO
  • Zong Yuntao
  • Wang Taosen

Assignees

  • 深圳市唯刚科技有限公司

Dates

Publication Date
20260512
Application Date
20260129

Claims (10)

  1. 1. The method for realizing the SaaS cloud service of the electronic commerce in the gold industry is characterized by comprising the following steps of: Collecting and extracting dimension characteristics of hyperspectral characteristic images, polarized characteristic images and defect mask images of gold products in real time to splice and form dimension-reducing fusion characteristic vectors; Based on the distribution situation of the dimension-reducing fusion feature vector in the low-dimension feature space, constructing a feature manifold reflecting the oxidation evolution continuity of the gold coating, selecting and calculating the local density entropy and the feature vector field divergence of gold coating sample points along an oxidation principal axis; Defining transition points of thickness variation of the gold plating layer as mutation points of local density entropy and characteristic vector field divergence, and automatically identifying a plurality of mutation points through a Bayesian mutation point detection algorithm so as to dynamically divide the order value grade and the core situation point; In the feature manifold space, calculating the geodesic distance between the feature vector of the dimension reduction fusion and the core situation point corresponding to each order level, and determining the order level of the gold product according to the situation interval in which the geodesic distance is located; Training through a pre-acquired reference coating reflectivity feature set, generating a reflectivity anomaly detection model, and inputting a dimension reduction fusion feature vector into the reflectivity anomaly detection model to output a reflectivity detection signal; And judging the qualified image grade of the gold product according to the grade value grade and the reflectivity detection signal.
  2. 2. The method for realizing the SaaS cloud service of the gold industry as claimed in claim 1, wherein the steps of constructing a feature manifold reflecting the continuity of oxidation evolution of the gold coating based on the distribution situation of the feature vector in the low-dimensional feature space by dimension reduction fusion, selecting and calculating the local density entropy and the feature vector field divergence of the gold coating sample point along the oxidation principal axis specifically comprise: mapping the dimension reduction fusion feature vector into a 2D low-dimensional space by adopting a nonlinear dimension reduction algorithm, extracting a spatial neighborhood relation of the oxidation state of the gold plating layer, constructing a local connection structure based on a K neighbor graph, and forming a feature manifold reflecting the oxidation evolution continuity of the gold plating layer by adopting a local linear embedding algorithm; Based on the distribution density of sample points in the feature manifold, automatically identifying a high-density region by using a density peak clustering algorithm and defining the high-density region as a core cluster of the oxidation state of the gold coating, and calculating and selecting the corresponding direction of the maximum feature value in the covariance matrix feature vector of the core cluster as an oxidation main shaft according to PCA; counting the number of sample points in the neighborhood of the radius R with each sample point as a center to obtain local density, calculating a probability density function of the local density through a kernel density estimation algorithm, and calculating local density entropy based on the probability density function and an information entropy formula; and taking each sample point in the characteristic manifold space as a base point, calculating the oxidation state change direction of K neighbor sample points, and constructing a characteristic vector field so as to calculate the divergence of the vector field at each sample point through a gradient estimation algorithm and obtain the characteristic vector field divergence.
  3. 3. The method for realizing the SaaS cloud service of the gold industry as claimed in claim 2, wherein after automatically identifying the high-density region by using a density peak clustering algorithm based on the distribution density of the sample points in the feature manifold and defining the high-density region as the core cluster of the oxidation state of the gold plating, calculating and selecting the corresponding direction of the maximum feature value in the covariance matrix feature vector of the core cluster as the oxidation main axis according to the PCA, further comprises: And carrying out alignment treatment on the oxygen body main axis directions of different samples by a dynamic time warping algorithm.
  4. 4. The method for realizing the SaaS cloud service of the gold industry electronic commerce according to claim 3, wherein the method is characterized in that the transition point of the thickness variation of the gold plating layer is defined as a mutation point of the local density entropy and the characteristic vector field divergence, and a plurality of mutation points are automatically identified through a bayesian mutation point detection algorithm so as to dynamically divide the level of the step value and the core situation point, and specifically comprises the following steps: Calculating the average value and standard deviation of local density entropy and characteristic vector field divergence in a window where each sample point is positioned by adopting a sliding window with the width of 5% -10% of the total number of samples and the step length of 1, so as to mark the sample points with the local density entropy exceeding the sum of the current window average value and 2 times of standard deviation and the characteristic vector field divergence exceeding the sum of the current window average value and 1.5 times of standard deviation as mutation candidate points, and carrying out iterative calculation on the positions and the number of the mutation candidate points by using a Bayesian mutation point detection algorithm to automatically identify a plurality of mutation points; If the number of the mutation points is N, dividing the characteristic manifold into N+1 oxidation stage areas according to the N mutation points, and performing Z-score standardization on the local density entropy and the characteristic vector field divergence of the sample points in each oxidation stage area so as to divide the oxidation state into a thin oxygen order value, a medium oxygen order value and a thick oxygen order value by adopting a K-means clustering algorithm; Calculating the local density of sample points in each oxidation stage area based on a DBSCAN algorithm, sorting, selecting sample points with the local density of 10% at the front as candidate points, calculating the cosine value of the included angle between each candidate point and the divergence direction of 5 nearest neighbor sample points, and selecting each candidate point with the included angle cosine value of more than 0.8 and the number of more than 3 as a core situation point.
  5. 5. The method for realizing the SaaS cloud service of the gold industry as defined in claim 4, wherein in the feature manifold space, a geodesic distance between the feature vector of the dimension reduction fusion and a core situation point corresponding to each level is calculated, and the level of the gold product is determined according to the situation interval in which the geodesic distance is located, specifically comprising: Mapping all the output core situation points into a feature manifold space and acquiring coordinates of each core situation point; Calculating the distance between each sample point in all sample points and each core situation point with the shortest distance from each sample point based on Dijkstra algorithm, and outputting the distance as a geodesic distance; dividing the characteristic manifold space into a thin oxygen interval, a medium oxygen interval and a thick oxygen interval by adopting a quantitive method according to the geodesic distance distribution of all the core situation points; And matching the corresponding rank value grades based on the interval of the geodesic distance of each sample point so as to determine the rank value grade of the gold product.
  6. 6. The method for realizing the SaaS cloud service of the gold industry as defined in claim 5, wherein before the feature manifold space is divided into the thin oxygen interval, the medium oxygen interval and the thick oxygen interval by adopting the fractional number method according to the geodesic distance distribution of all the core situation points, the method further comprises: And taking the initial geodesic distance of each sample point as a weight, selecting the sample points in the radius R neighborhood of the sample points to carry out local weighted regression processing, and correcting the geodesic distance through iterative calculation.
  7. 7. The method for implementing the SaaS cloud service of the gold industry as defined in claim 6, wherein the matching the corresponding rank value grades based on the interval of the geodesic distance of each sample point to determine the rank value grade of the gold product further comprises: if the geodesic distance of any sample point is at the boundary of two intervals, the nearest neighbor rule voting is adopted to determine the rank value.
  8. 8. The method for realizing the SaaS cloud service by the golden industry electronic commerce according to claim 7, wherein the training by the pre-acquired reference coating reflectivity feature set and generating a reflectivity anomaly detection model, inputting a dimension-reduction fusion feature vector into the reflectivity anomaly detection model to output a reflectivity detection signal, specifically comprises the following steps: measuring the reflectivity value of the gold plating layer by using a spectrophotometer while splicing to form a dimension-reducing fusion feature vector, and acquiring a reference plating layer reflectivity feature set by combining the dimension-reducing fusion feature vector and the reflectivity value; Taking a reference coating reflectivity feature set as input, training through an isolated forest algorithm, and generating a reflectivity anomaly detection model; inputting the dimension reduction fusion feature vector into a reflectivity anomaly detection model to output a reflectivity detection signal; and judging the qualified image grade of the gold product according to the grade of the order value and the reflectivity detection signal.
  9. 9. The method for realizing the SaaS cloud service of the gold industry electronic commerce according to claim 8, wherein the judging of the image qualified grade of the gold product according to the order grade and the reflectivity detection signal comprises the following steps: to output a high grade and execute an image processing termination command when a thin oxygen level value and a normal reflectivity detection signal are input; when a thin oxygen level value and an abnormal reflectivity detection signal or a medium oxygen level value and a normal reflectivity detection signal are input, outputting a medium grade and executing an image rendering material parameter correction command; when a medium oxygen order value and an abnormal reflectivity detection signal or a thick oxygen order value and a normal reflectivity detection signal are input, outputting a medium lower grade and executing an oxide layer distribution area material replacement command; When the thick oxygen level value and the abnormal reflectivity detection signal are input, a low-grade image rendering material parameter correction command and an oxide layer distribution area material replacement command are output and executed simultaneously.
  10. 10. The gold industry e-commerce SaaS cloud service implementation system is characterized by being used for implementing the gold industry e-commerce SaaS cloud service implementation method according to any one of the claims 1-9, and the gold industry e-commerce SaaS cloud service implementation system comprises: The feature acquisition module is used for acquiring and extracting dimension features of the hyperspectral feature image, the polarization feature image and the defect mask image of the gold product in real time so as to splice and form a dimension-reducing fusion feature vector; the sample parameter calculation module can construct a characteristic manifold reflecting the oxidation evolution continuity of the gold coating based on the distribution situation of the dimension-reduction fusion characteristic vector in the low-dimension characteristic space, and select and calculate the local density entropy and the characteristic vector field divergence of gold coating sample points along the oxidation principal axis; the step value grade dividing module is used for defining the transition point of the thickness change of the gold plating layer as a mutation point of the local density entropy and the characteristic vector field divergence, and automatically identifying a plurality of mutation points through a Bayesian change point detection algorithm so as to dynamically divide the step value grade and the core situation point; The step value grade determining module is used for calculating the geodesic distance between the dimension-reduction fusion feature vector and the core situation point corresponding to each step value grade in the feature manifold space, and determining the step value grade of the gold product according to the situation interval in which the geodesic distance is located; The reflectivity detection signal output module can train and generate a reflectivity anomaly detection model through a pre-acquired reference coating reflectivity feature set, and input a dimension reduction fusion feature vector into the reflectivity anomaly detection model to output a reflectivity detection signal; And judging the qualified image grade of the gold product according to the grade value grade and the reflectivity detection signal.

Description

Gold industry e-commerce SaaS cloud service realization method and system Technical Field The invention relates to the technical field of electronic commerce services in the gold industry, in particular to a method and a system for realizing electronic commerce SaaS cloud services in the gold industry. Background The gold industry electronic commerce SaaS cloud service is a whole-flow electronic business management system formed by links of image processing, marketing, distribution and the like of gold. However, the gold surface is easy to form a double-layer dielectric film structure composed of a micron-sized oxide layer and an electroplated layer due to environmental factors, and the existence of the double-layer dielectric film structure leads to the fact that the obtained gold image is easy to show non-uniform iridescent diffraction speckles, so that the reduction of the natural color of gold is seriously damaged. Although the existing gold industry electronic commerce SaaS cloud service implementation method has been used for realizing the improvement of image quality through operations such as image rendering material parameter correction or oxide layer distribution area material replacement, the space-time difference between the thickness distribution of the oxide layer on the gold surface and the reflectivity of the coating is difficult to analyze, so that consumers cannot truly observe the color and texture of gold products through images, and the development of gold industry electronic commerce is severely restricted. Therefore, how to analyze the space-time difference between the thickness distribution of the oxide layer on the gold surface and the reflectivity of the plating layer becomes a considerable problem for the person skilled in the art. Disclosure of Invention The invention aims to solve the problems that the prior method for realizing the SaaS cloud service of the electronic commerce in the gold industry is difficult to analyze the space-time difference between the thickness distribution of the oxide layer on the gold surface and the reflectivity of the plating layer, so that consumers cannot truly observe the color and texture of gold products through images, and the development of the electronic commerce in the gold industry is seriously restricted. In order to achieve the purpose, the invention provides a method and a system for realizing the SaaS cloud service of the electronic commerce in the gold industry. According to a first aspect of the invention, a method for realizing the SaaS cloud service of the electronic commerce in the gold industry is provided, which comprises the following steps: Collecting and extracting dimension characteristics of hyperspectral characteristic images, polarized characteristic images and defect mask images of gold products in real time to splice and form dimension-reducing fusion characteristic vectors; Based on the distribution situation of the dimension-reducing fusion feature vector in the low-dimension feature space, constructing a feature manifold reflecting the oxidation evolution continuity of the gold coating, selecting and calculating the local density entropy and the feature vector field divergence of gold coating sample points along an oxidation principal axis; Defining transition points of thickness variation of the gold plating layer as mutation points of local density entropy and characteristic vector field divergence, and automatically identifying a plurality of mutation points through a Bayesian mutation point detection algorithm so as to dynamically divide the order value grade and the core situation point; In the feature manifold space, calculating the geodesic distance between the feature vector of the dimension reduction fusion and the core situation point corresponding to each order level, and determining the order level of the gold product according to the situation interval in which the geodesic distance is located; Training through a pre-acquired reference coating reflectivity feature set, generating a reflectivity anomaly detection model, and inputting a dimension reduction fusion feature vector into the reflectivity anomaly detection model to output a reflectivity detection signal; And judging the qualified image grade of the gold product according to the grade value grade and the reflectivity detection signal. Optionally, based on the distribution situation of the dimension-reducing fusion feature vector in the low-dimensional feature space, constructing a feature manifold reflecting the oxidation evolution continuity of the gold coating, selecting and calculating the local density entropy and the feature vector field divergence of the gold coating sample points along the oxidation principal axis, wherein the method specifically comprises the following steps: mapping the dimension reduction fusion feature vector into a 2D low-dimensional space by adopting a nonlinear dimension reduction algorithm, extracting a spatial nei