Search

CN-121997174-A - Method and system for detecting heavy metal pollution of common mussel based on CNN and femtosecond laser-induced breakdown spectroscopy

CN121997174ACN 121997174 ACN121997174 ACN 121997174ACN-121997174-A

Abstract

The invention belongs to the technical field of aquatic food safety detection, and in particular relates to a method and a system for detecting heavy metal pollution of a Mytilus edulis based on CNN and femtosecond laser-induced breakdown spectroscopy, which are realized based on an Fs-LA-LIBS spectrum detection device, wherein the method comprises the steps of preparing a sample of the Mytilus edulis to be detected, and placing the sample on a three-dimensional translation stage of the Fs-LA-LIBS spectrum detection device; the method comprises the steps of carrying out spectrum collection and pretreatment on a sample based on an Fs-LA-LIBS spectrum detection device to obtain statistical characteristics, constructing a Mytilus edulis heavy metal pollution detection model based on a CNN network model, and obtaining a Mytilus edulis heavy metal pollution detection result based on the statistical characteristics and the Mytilus edulis heavy metal pollution detection model. According to the invention, through the embedded collaborative architecture of CNN and femtosecond laser-induced breakdown spectroscopy, three technical problems existing in the detection of the heavy metal of the common mussel in the traditional method are solved, and the nondestructive, high-precision and online quantitative detection is realized.

Inventors

  • HE XIAOYONG
  • Chen Xuemian
  • GAO ZHENMAN

Assignees

  • 东莞理工学院

Dates

Publication Date
20260508
Application Date
20260127

Claims (9)

  1. 1. The method for detecting the heavy metal pollution of the common mussel based on the CNN and the femtosecond laser induced breakdown spectroscopy is characterized by being realized based on an Fs-LA-LIBS spectrum detection device, and comprises the following steps: preparing a sample from the Mytilus edulis to be detected, and placing the sample on a three-dimensional translation stage of an Fs-LA-LIBS spectrum detection device; carrying out spectrum acquisition and pretreatment on a sample based on an Fs-LA-LIBS spectrum detection device to obtain statistical characteristics; Constructing a Mytilus edulis heavy metal pollution detection model based on the CNN network model; based on the statistical characteristics and the perna canaliculus heavy metal pollution detection model, a perna canaliculus heavy metal pollution detection result is obtained.
  2. 2. The method of claim 1, wherein the Fs-LA-LIBS spectrum detection device comprises a femtosecond laser, a fiber spectrometer, a three-dimensional translation stage and a direct-current high-voltage power supply; Femtosecond laser, namely using a Ti-sapphire laser as an excitation light source for LIBS experiments, wherein the central wavelength is 800nm, the pulse width is 50fs, and the frequency is 1kHz; the optical fiber spectrometer has a wavelength range of 200-550nm and a resolution of 0.07nm; The three-dimensional translation stage is driven by a stepping motor; DC high-voltage power supply with voltage of 10kV and current of 0.2A.
  3. 3. The method of claim 1, wherein the method for obtaining statistical features by performing spectral collection and preprocessing on the sample based on the Fs-LA-LIBS spectral detection device comprises: carrying out spectrum acquisition on a sample by adopting a space multipoint sampling and time sequence averaging method to obtain spectrum data; normalizing the spectrum data to obtain preprocessed spectrum data; Extracting peak characteristics of the preprocessed spectrum data by adopting a self-adaptive threshold method to obtain key indexes, wherein the key indexes comprise peak value number, peak value intensity mean value and peak value position deviation; and calculating the spectrum integral intensity mean value and the intensity standard deviation based on the key indexes to obtain statistical characteristics, wherein the statistical characteristics comprise peak value number, peak value intensity mean value, integral intensity mean value and intensity standard deviation.
  4. 4. A method according to claim 3, wherein the method for obtaining spectral data by performing spectral acquisition on the sample using a spatial multi-point sampling + time series averaging method comprises: The three-dimensional translation stage is used for controlling the sample stage to move step by step along the X/Y axis, 5 detection points which are uniformly distributed are selected on the surface of the sample, a 3X 3 grid center and a vertex are formed, each detection point is used for collecting 3 spectra, and the influence of single-point abnormal values is avoided; the single spectrum acquisition time length is set to be 100ms, plasma flicker noise is reduced through an average algorithm built in the optical fiber spectrometer, each sample outputs 5 groups of spectrum data, and each group contains an average value of 3 repeated acquisitions.
  5. 5. The method of claim 1, wherein the method for obtaining the detection result of the heavy metal contamination of the mytilus edulis based on the statistical characteristics and the detection model of the heavy metal contamination of the mytilus edulis comprises: The convolution feature extraction layer of the perna canaliculus heavy metal pollution detection model aims at the 1-dimensional sequence characteristic of spectrum data wavelength-intensity, and 3 stacked convolution blocks are adopted to realize feature extraction, each convolution block comprises 2 layers of 1-dimensional convolution operation, convolution kernel size is 3, reLU activation functions are normalized and maximally pooled, and the pooled kernel size is 2, and the shape, width and peak interval information of heavy metal feature peaks are automatically identified through the convolution feature extraction layer to obtain 1024-dimensional output features; The characteristic fusion layer of the perna canaliculus heavy metal pollution detection model performs channel dimension splicing on 1024-dimensional output characteristics output by the convolution characteristic extraction layer and 4-dimensional statistical characteristics to form 1028-dimensional composite characteristic vectors; The full-connection layer of the Mytilus edulis heavy metal pollution detection model adopts a 3-layer full-connection network to construct a classifier, and is matched with a Dropout regularization inhibition model to be fitted, the first two layers of full-connection networks gradually compress 1028-dimensional composite features to 256-dimensional and 128-dimensional, core discrimination information is reserved, the last one layer of full-connection network maps the 128-dimensional features to 3-dimensional output, the 3-dimensional output is converted into probability values of all preset levels through a Softmax function, and the level with the maximum probability value is taken as a Mytilus edulis heavy metal pollution detection result.
  6. 6. The method of claim 1, further comprising training a perna canaliculus heavy metal contamination detection model: Collecting data of four types of Mytilus edulis samples with different concentrations, dividing the data into a training set, a verification set and a test set according to preset proportions, and training a Mytilus edulis heavy metal pollution detection model based on the training set, the verification set and the test set; PCA or t-SNE is adopted in the training process to realize the dimension reduction visualization of the high-dimension characteristic of the heavy metal spectrum of the perna canaliculus; in the training process, an adaptive moment estimation optimizer is adopted to combine with a cosine annealing learning rate scheduling strategy; and in the training process, gaussian noise with the signal-to-noise ratio of 20dB is injected into the test set, and noise interference existing in actual detection is simulated.
  7. 7. The system for detecting the heavy metal pollution of the common mussel based on CNN and femtosecond laser-induced breakdown spectroscopy is used for realizing the method of any one of claims 1 to 6, and is characterized by comprising a sample preparation module, a spectrum acquisition processing module, a model construction module and a detection module; The sample preparation module is used for preparing the Mytilus edulis to be detected into a sample and placing the sample on a three-dimensional translation stage of the Fs-LA-LIBS spectrum detection device; The spectrum acquisition processing module is used for carrying out spectrum acquisition and pretreatment on the sample based on the Fs-LA-LIBS spectrum detection device to obtain statistical characteristics; the model construction module is used for constructing a Mytilus edulis heavy metal pollution detection model based on the CNN network model; And the detection module is used for obtaining the detection result of the heavy metal pollution of the mytilus edulis based on the statistical characteristics and the detection model of the heavy metal pollution of the mytilus edulis.
  8. 8. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the method of any one of claims 1-6 when the program is executed.
  9. 9. A computer readable storage medium, characterized in that the computer readable storage medium stores a computer program, which, when executed, implements the method of any of claims 1-6.

Description

Method and system for detecting heavy metal pollution of common mussel based on CNN and femtosecond laser-induced breakdown spectroscopy Technical Field The invention belongs to the technical field of aquatic food safety detection, and particularly relates to a method and a system for detecting heavy metal pollution of common mussels based on CNN and femtosecond laser-induced breakdown spectroscopy. Background The common mussel is used as an offshore pollution indicator, and the heavy metal enrichment characteristic of the common mussel is widely applied to marine environment monitoring. The current heavy metal detection technology for the Mytilus edulis mainly comprises a traditional laboratory detection method and a LIBS combined machine learning scheme, and has obvious limitations. The following details the closest technical solutions and their drawbacks: 1. The traditional laboratory detection method mainly comprises an Atomic Absorption Spectrometry (AAS) or an inductively coupled plasma mass spectrometry (ICP-MS) and the like, wherein the AAS is based on the absorption of ground state atoms to light with specific wavelength, the concentration of elements is quantified through absorbance, the ICP-MS is separated according to mass-to-charge ratio (m/z) after ionization of a sample, and the elements are quantified through ion flow intensity, and the two methods can detect the heavy metal content of the Mytilus edulis, but have the defects that (1) the sample is destroyed, the AAS and the ICP-MS both need strong acid to digest the Mytilus edulis sample, the biological structure is thoroughly destroyed, the spatial distribution of the heavy metal and the chemical morphology information are permanently lost, and the same living body cannot be retested, and (2) the cost is high, the mass of the sample required for single detection exceeds 50g, the chemical agent is required for detection, the time is about 2 hours, and the timeliness is low. LIBS-coupled machine learning scheme in which Random Forest (RF), support Vector Machine (SVM), and partial least squares analysis (PLS) are the dominant approaches. The random forest improves classification robustness through multi-decision tree integration and feature importance sorting, but depends on manual screening features, nonlinear interaction (such as Cd/Pb spectral line overlapping) in a high-dimensional spectrum is difficult to adaptively decouple, a support vector machine processes nonlinear classification by utilizing kernel function mapping, is sensitive to noise (boundary failure in time division caused by feature deviation due to plasma flicker) and cannot output concentration classification results, and partial least square analysis solves the problem of collinearity through linear dimension reduction, but cannot capture complex response relation between a Mytilus edulis organic matrix and heavy metals (such as masking effect of Ca matrix on Cd signals). The common defects of the three are that the characteristic engineering relies on experience and has poor noise robustness, and only qualitative judgment is supported, so that the quantitative classification requirement of heavy metals can not be met. Therefore, a new detection method for heavy metal pollution of the perna canaliculus is needed to realize nondestructive, high-precision and online quantitative detection. Disclosure of Invention In order to solve the problems in the prior art, the invention provides a method and a system for detecting heavy metal pollution of the common mussel based on CNN and femtosecond laser induced breakdown spectroscopy, and aims to solve the three technical problems of sample damage and dynamic monitoring failure caused by strong acid digestion in the traditional method in heavy metal detection of the common mussel, misjudgment (such as Cd-Ca overlapping) of trace element spectral lines caused by artificial characteristic dependence in the traditional LIBS intelligent scheme, and incapability of outputting pollution classification in real time, thereby finally realizing nondestructive, high-precision and online quantitative detection. In order to achieve the above object, the present invention provides the following solutions: The method for detecting the heavy metal pollution of the common mussel based on the CNN and the femtosecond laser induced breakdown spectroscopy is realized based on an Fs-LA-LIBS spectrum detection device, and comprises the following steps: preparing a sample from the Mytilus edulis to be detected, and placing the sample on a three-dimensional translation stage of an Fs-LA-LIBS spectrum detection device; carrying out spectrum acquisition and pretreatment on a sample based on an Fs-LA-LIBS spectrum detection device to obtain statistical characteristics; Constructing a Mytilus edulis heavy metal pollution detection model based on the CNN network model; based on the statistical characteristics and the perna canaliculus heavy metal pollution det