Intelligent detection method for purity of gold jewelry and ornament

Abstract

The invention relates to the field of ornament purity detection, in particular to an intelligent gold ornament purity detection method which comprises the following steps: acquiring a spectral data sequence and an ornament label of a target ornament in history under multiple spectrums, and calculating the distinction degree of the target ornament; in response to the condition that the distinction degree is not smaller than a preset distinction threshold value, inputting the distinction degree into a first detection network to output a detection result of the target ornament; in response to the condition that the distinction degree is smaller than a preset distinguishing threshold value, exciting the preprocessed target ornament according to a preset excitation signal to obtain a vibration signal of the target ornament, performing spectral analysis on the vibration signal to obtain a resonant frequency variation, and inputting the distinction degree and the resonant frequency variation into a second detection network to output a detection result of the target ornament; and completing purity detection according to a detection result. According to the technical scheme, the precision and efficiency of the purity detection result of the ornament can be improved.

Description

Technical Field

[0001] The present invention relates to the field of jewelry purity detection, and more specifically, to an intelligent method for detecting the purity of gold jewelry. Background Art

[0002] With the increasing demand for jewelry, especially gold jewelry, the purchase, trading, and collection of gold jewelry are becoming increasingly frequent. Accurately testing the purity of gold jewelry to ensure it meets standards and consumer expectations has become a key issue during the quality assessment and trading process. Traditional methods for testing the purity of gold jewelry typically rely on manual sampling, chemical analysis, or high-precision instruments and equipment. These methods are not only time-consuming but also often affected by environmental and operational factors, resulting in significant errors in the test results. Summary of the Invention

[0003] To address the technical problem of inaccurate intelligent purity detection results for gold jewelry due to traditional methods relying on a single detection method, the present invention provides an intelligent purity detection method for gold jewelry. The method comprises: obtaining a historical multi-spectral spectral data sequence and a jewelry label for a target jewelry item, where the target jewelry item is a jewelry item with any type of coating, and the jewelry label is the coating type; calculating the discrimination of the target jewelry item; in response to the discrimination being no less than a preset discrimination threshold, inputting the discrimination into a first detection network to output a detection result for the target jewelry item, the detection result being a probability of meeting the purity standard and a probability of failing the purity standard; in response to the discrimination being less than the preset discrimination threshold, exciting the pre-processed target jewelry item with a preset excitation signal to obtain a vibration signal of the target jewelry item, performing spectral analysis on the vibration signal to obtain a resonant frequency change, inputting the discrimination and resonant frequency change into a second detection network to output a detection result for the target jewelry item; and completing purity detection based on the detection result.

[0004] Preferably, the multispectral analysis includes: XRF or LIBS.

[0005] Preferably, the discrimination includes: taking any type of jewelry with a coating other than the target jewelry as a reference jewelry; respectively calculating the similarity between the spectral data sequence of the target jewelry and the spectral data sequence of any reference jewelry, and using the mean of all similarities through negative correlation mapping as the discrimination.

[0006] Preferably, the discrimination further comprises: taking any ornament of a coating type other than the target ornament as a reference ornament; the discrimination satisfies the relationship:

[0007] s i represents the discrimination of target ornament i, f i represents the spectral data sequence of target jewelry i, f jrepresents the spectral data sequence of reference ornament j, KL represents KL divergence, J represents the total number of parameter ornaments, and exp represents the exponential function.

[0008] Preferably, the first detection network is a network model constructed with the spectral data sequence of historical jewelry as input, the jewelry label as the network label, and the detection result as output.

[0009] Preferably, the preprocessing includes removing noise, detrending, and eliminating the influence of contact resistance through differential electrodes to ensure the accuracy of the measurement signal.

[0010] Preferably, the spectrum analysis of the vibration signal to obtain the resonant frequency change includes: performing Fourier transform on the vibration signal, converting the time domain signal into a frequency domain signal, and extracting the frequency of the resonant peak; comparing the resonant frequencies of different sweeping stages, and calculating the difference in resonant frequencies between two adjacent sweeping stages as the frequency change.

[0011] Preferably, the second detection network is a CNN network.

[0012] Beneficial effects of the present invention:

[0013] The present invention effectively improves the accuracy and efficiency of gold jewelry purity detection by combining multispectral data with spectral analysis of vibration signals. First, the spectral data of the jewelry is acquired through multispectral analysis, and the jewelry with different coating types are distinguished by calculating the discrimination degree, thereby providing a reliable basis for subsequent purity detection. When the discrimination degree is high, a deep learning network is used for direct purity detection; when the discrimination degree is low, the vibration signal is generated by stimulating the jewelry and analyzing the change in its resonant frequency, thereby further improving the accuracy and sensitivity of the detection. This method can not only eliminate interference factors such as contact resistance, but also further conduct in-depth analysis of the physical properties of the jewelry through the frequency change, providing a more comprehensive purity judgment. The present invention uses multimodal data processing and deep learning technology to greatly enhance the robustness and automation level of gold jewelry purity detection, and can achieve efficient and accurate purity detection in practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 The present invention is a flowchart of an intelligent method for detecting the purity of gold jewelry. DETAILED DESCRIPTION

[0015] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0016] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0017] Reference Figure 1 A method for intelligently detecting the purity of gold jewelry includes steps S1 to S3, specifically as follows:

[0018] S1: Obtain the spectral data sequence and jewelry label of the target jewelry under multi-spectral conditions in history, and calculate the discrimination degree of the target jewelry.

[0019] It should be noted that to study the relationship between the target jewelry's multispectral spectral data sequence and its coating type, spectral data for jewelry with any coating type was acquired. Specifically, multispectral imaging technology was used to collect the target jewelry's reflectance spectral data in different spectral bands, forming a complete set of spectral data sequences. Furthermore, to further analyze the characteristics of each coating type, each jewelry piece was assigned a label indicating its corresponding coating type. By comparing the spectral data of different coating types, the discrimination of the target jewelry pieces was calculated. Specifically, by analyzing their differences in multiple spectral bands, the differences in the spectral characteristics of jewelry with different coating types were evaluated.

[0020] In one embodiment, X-ray fluorescence (XRF) and laser-induced breakdown spectroscopy (LIBS) are employed to accurately acquire multispectral spectral data sequences for target jewelry. Both techniques are effective for analyzing the elemental composition and chemical characteristics of jewelry surfaces and their coatings. XRF excites surface elements on the sample, causing them to emit characteristic X-rays, thereby acquiring elemental analysis data. This is particularly suitable for analyzing the types and content of elements in coatings. LIBS, on the other hand, uses laser pulses to excite the surface material of jewelry, causing it to ionize and excite, thereby acquiring spectral data. This enables rapid, high-resolution surface elemental analysis and is particularly suitable for testing complex samples. These two methods can extract detailed information about jewelry from different spectral signatures, label its coating type, and further analyze and calculate the spectral differentiation between jewelry with different coating types. The combination of XRF and LIBS not only improves detection accuracy but also enhances the ability to distinguish different coating types. This combination of methods effectively overcomes the limitations of traditional spectral methods, achieving higher differentiation and reliability, especially when the sample surface is complex and variable.

[0021] Select any jewelry with a coating type other than the target jewelry as a reference jewelry. Using multispectral data analysis techniques, calculate the similarity between the target jewelry's spectral data sequence and the spectral data sequence of each reference jewelry. Average these similarities to obtain the mean similarity between all reference jewelry and the target jewelry. To further quantify discrimination, transform these mean similarity values using a negative correlation mapping to obtain the final discrimination value. The Pearson correlation coefficient can be used as the similarity measure.

[0022] Negative correlation mapping can effectively reflect the inverse relationship of similarity, ensuring that the lower the similarity, the higher the discrimination, thereby effectively identifying the difference between the target ornament and the reference ornament.

[0023] The benefits of this approach are: first, by comparing with multiple reference jewelry, it enhances the comprehensiveness and reliability of discrimination calculations, avoiding the potential bias associated with relying solely on a single reference jewelry piece. Second, negative correlation mapping enables a more intuitive and accurate assessment of the differences between the target jewelry piece and different coating types, thereby improving jewelry classification and recognition accuracy.

[0024] In another embodiment, any type of ornament with a coating other than the target ornament is used as a reference ornament. The discrimination degree satisfies the relationship:

[0025] s i represents the discrimination of target ornament i, f i represents the spectral data sequence of target jewelry i, f j represents the spectral data sequence of reference ornament j, KL represents KL divergence, J represents the total number of parameter ornaments, and exp represents the exponential function.

[0026] The KL divergence is used to quantify the difference in spectral characteristics between the two. By introducing an exponential function to map similarity, the difference between the target and reference jewelry can be effectively enhanced, and the uniqueness of the target jewelry can be evaluated by calculating the discrimination degree. The core physical significance of this method lies in revealing the differences in the spectral characteristics of the target jewelry by comparing the spectral data of multiple reference jewelry, thereby improving the ability to distinguish the target jewelry from other jewelry. In practical applications, this method has high practical value because it can accurately extract and identify the unique characteristics of the target jewelry in a complex jewelry sample, enhancing the robustness and accuracy of the analysis.

[0027] S2: Complete purity testing based on the test results.

[0028] In one implementation, the discrimination is compared with a preset discrimination threshold. If the discrimination is not less than the preset discrimination threshold, the discrimination is input into a first detection network, which then outputs a detection result for the target jewelry. This result includes a probability of meeting the purity standard and a probability of failing the purity standard, indicating whether the target jewelry meets the standard purity requirements. The first detection network is a deep learning model trained based on historical jewelry spectral data sequences and corresponding jewelry labels. In this way, the network learns the relationship between the spectral characteristics of different jewelry and their purity, effectively assessing the purity of the target jewelry.

[0029] Combining traditional spectral analysis with deep learning technology, the system compares discrimination against preset thresholds to ensure that only target jewelry with distinct distinguishing features is processed further. This layered detection process improves the system's efficiency and accuracy, avoiding unnecessary false or missed detections. Furthermore, the network model trained using historical data captures a wider range of jewelry characteristics, making the system adaptable to different types of jewelry and further enhancing the reliability and applicability of purity testing.

[0030] In response to the degree of discrimination being less than a preset discrimination threshold, the preprocessed target ornament is excited according to the preset excitation signal to obtain a vibration signal of the target ornament, the vibration signal is subjected to spectral analysis to obtain a change in the resonant frequency, and the degree of discrimination and the change in the resonant frequency are input into a second detection network to output a detection result of the target ornament.

[0031] Specifically, noise removal is required to reduce the impact of external interference on the measured signal and ensure a true reflection of the signal. Detrending is the second step, aiming to eliminate trend changes in the data caused by factors such as systematic errors or time variations, so that the signal can more accurately reflect the true characteristics of the target signal. To further improve measurement accuracy, differential electrode technology is used to eliminate the influence of contact resistance. Contact resistance can cause signal distortion due to poor contact between the electrode and the jewelry. Differential electrodes can eliminate the influence of this resistive error by simultaneously measuring the current difference between the positive and negative electrodes, making the final measurement results more accurate and reliable.

[0032] The pre-processed target jewelry is excited using a preset excitation signal to obtain a vibration signal. The vibration signal is then spectrally analyzed to determine the change in resonant frequency. Excitation is achieved by outputting a sweep signal from 1kHz to 100kHz using a sweep signal generator. The sweep process involves gradually varying the signal within this frequency range (typically a linear or logarithmic sweep) for a specific duration. This signal is then applied to the jewelry sample via an excitation device. The excitation device should select an appropriate excitation method based on the geometry of the jewelry sample.

[0033] The Fourier transform of the collected vibration signal is performed to convert the time-domain signal into the frequency domain. Typically, a discrete Fourier transform (DFT) or fast Fourier transform (FFT) is used. This effectively reveals the energy distribution of the signal at different frequencies. After obtaining the frequency-domain signal, the characteristic frequency of the vibration system is identified by extracting the frequency of the resonant peak. The resonant frequency changes during different sweep stages are compared, and the difference between the resonant frequencies of two adjacent sweep stages is calculated as the frequency variation.

[0034] The discrimination degree and the change in the resonant frequency are input into the second detection network to output the detection result of the target jewelry. The second detection network is a CNN network.

[0035] Complete purity testing based on the test results.

[0036] It should be noted that those skilled in the art may make various modifications and improvements without departing from the scope of the present invention, and these modifications and improvements fall within the scope of protection of the present invention. Therefore, the scope of protection of the patent for this invention shall be based on the appended claims.

Claims

1. An intelligent method for detecting the purity of gold jewelry, characterized in that: include: Obtain the spectral data sequence and label of the target jewelry under multi-spectral conditions in the history. The target jewelry is a jewelry with any coating type, and the label is the coating type. Calculate the discrimination of the target jewelry. In response to the discrimination being no less than a preset discrimination threshold, the discrimination is input into a first detection network to output a detection result of the target jewelry, the detection result being a probability of the purity meeting the standard and a probability of the purity failing to meet the standard; in response to the discrimination being less than the preset discrimination threshold, the pre-processed target jewelry is excited according to a preset excitation signal to obtain a vibration signal of the target jewelry, a spectrum analysis is performed on the vibration signal to obtain a change in the resonant frequency, and the discrimination and the change in the resonant frequency are input into a second detection network to output a detection result of the target jewelry; Complete purity testing based on the test results.

2. The method for intelligently detecting the purity of gold jewelry according to claim 1, characterized in that: The multi-spectrum includes: XRF and LIBS.

3. The method for intelligently detecting the purity of gold jewelry according to claim 1, characterized in that: The discrimination includes: Use any accessory with a plating type other than the target accessory as a reference accessory; The similarity between the spectral data sequence of the target jewelry and the spectral data sequence of any reference jewelry is calculated respectively, and the mean of all similarities is used as the discrimination result through negative correlation mapping.

4. The method for intelligently detecting the purity of gold jewelry according to claim 1, characterized in that: The discrimination also includes: Use any accessory with a plating type other than the target accessory as a reference accessory; The discrimination satisfies the relationship: s i represents the discrimination of target ornament i, f i represents the spectral data sequence of target jewelry i, f j represents the spectral data sequence of reference ornament j, KL represents KL divergence, J represents the total number of parameter ornaments, and exp represents the exponential function.

5. The method for intelligently detecting the purity of gold jewelry according to claim 1, characterized in that: The first detection network is a network model constructed with the spectral data sequence of historical jewelry as input, the jewelry label as the network label, and the detection result as output.

6. The method for intelligently detecting the purity of gold jewelry according to claim 1, characterized in that: The preprocessing includes removing noise, detrending, and eliminating the influence of contact resistance through differential electrodes to ensure the accuracy of the measurement signal.

7. The method for intelligently detecting the purity of gold jewelry according to claim 1, characterized in that: The performing spectrum analysis on the vibration signal to obtain the resonant frequency variation comprises: Perform Fourier transform on the vibration signal to convert the time domain signal into the frequency domain signal and extract the frequency of the resonance peak; The resonant frequencies in different sweep stages are compared, and the difference in resonant frequencies between two adjacent sweep stages is calculated as the frequency variation.

8. The method for intelligently detecting the purity of gold jewelry according to claim 1, characterized in that: The second detection network is a CNN network.

Images