Microplastic quantity and type intelligent identification method based on machine vision learning

Abstract

The invention discloses a machine vision learning-based micro-plastic quantity and type intelligent identification method, and relates to the technical field of image identification and micro-plastic detection. The method at least comprises the following steps: S1, obtaining quantity and morphological characteristic information of micro-plastics through an experiment or other related methods, and combining and processing to form a typical characteristic database of the micro-plastics; s2, introducing machine learning and a YOLO algorithm to construct a basic recognition model; on the basis of machine vision learning, an efficient, rapid, convenient-to-operate and low-cost intelligent quantitative method for the number of the micro-plastics is established, the detection time is greatly shortened, the types of the micro-plastics can be distinguished by identifying the colors of the micro-plastics, the problem that the micro-plastics are difficult to analyze on a large scale in an existing method can be solved, and the method is suitable for large-scale analysis of the micro-plastics. And an accurate, efficient and economical micro-plastic intelligent batch identification effect is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition and microplastic detection, and specifically to a method for intelligently identifying the quantity and type of microplastics based on machine vision learning. Background Art

[0002] At present, the detection of microplastics is mainly based on optical and chemical methods. The stereo microscope based on the optical method can identify microplastics larger than 1 mm, while the scanning electron microscope can analyze the surface of nano-scale microplastic particle samples to obtain the surface morphology of the samples. At the same time, Fourier transform infrared spectroscopy can perform non-destructive qualitative and quantitative analysis on microplastic particles > 50 μm, while Raman spectroscopy can perform non-destructive qualitative and quantitative analysis on microplastic particles < 50 μm. Thermal desorption-gas chromatography-mass spectrometry and pyrolysis-gas chromatography-mass spectrometry based on chemical methods can accurately characterize and quantify single microplastic particle samples and analyze the specific additives of microplastics.

[0003] Existing microplastic detection technologies have disadvantages such as expensive equipment, time-consuming and labor-intensive, low detection efficiency, inability to process samples in batches, and complex detection process.

[0004] Therefore, a new solution to the above problems needs to be proposed. Summary of the invention

[0005] The purpose of the present invention is to provide an intelligent identification method for the quantity and type of microplastics based on machine vision learning. By establishing an efficient, fast, easy-to-operate, and low-cost intelligent quantitative identification method for the quantity of microplastics based on machine vision learning, the detection time is greatly reduced, and the types of microplastics can be distinguished by identifying the colors of the microplastics, so as to solve the technical problems raised in the background technology.

[0006] To achieve the above-mentioned purpose, the present invention provides the following technical solution: a method for intelligently identifying the quantity and type of microplastics based on machine vision learning, comprising at least the following steps:

[0007] S1: Obtain the quantity and morphological characteristics of microplastics through experiments and scientific literature, combine and process them to form a database of typical characteristics of microplastics;

[0008] S2: Introduce machine learning and YOLO algorithm to build a basic recognition model, input the chemical type and image information of the processed microplastics into the basic image recognition model for targeted training, and form an automatic recognition model for microplastics. The basic model for automatic recognition of plastics can accurately identify the chemical type and particle quantity information of images in the microplastics typical feature database, and has a certain recognition ability for new images;

[0009] S3: Acquire images of microplastics to be processed;

[0010] S4: Input the microplastic image to be processed into the automatic recognition model of microplastics, and obtain the recognition result of the image.

[0011] Furthermore, the step S2 of inputting the chemical property information and image information of the processed microplastics into the basic image recognition model for targeted training comprises at least the following steps:

[0012] Classify the morphological features of microplastics in the training images to obtain multiple groups of microplastic training images;

[0013] Based on the morphological features of microplastics in the training images, the basic image recognition model is trained to form an automatic microplastic recognition system.

[0014] Based on the method of constructing a typical feature database of microplastics, the color and quantity of microplastics are associated with images in the form of electronic tags to form microplastic training images;

[0015] The microplastic training images were divided into training set, test set and validation set in a ratio of 5:4:1 for basic recognition model training.

[0016] Furthermore, the introduction of machine learning and YOLO algorithm to construct a basic recognition model includes at least the following steps:

[0017] The model selection, test data type, prediction result saving path, input image size, confidence threshold and result display settings in the preset YOLO algorithm are used to improve the training speed of the basic model and the recognition speed of the microplastic automatic recognition model;

[0018] Select yolov8s as the model of the YOLO algorithm, the test data type is image, the input image resolution is set to 640×640 according to the computer parameters, and the confidence threshold is set to 0.80 by default;

[0019] Cut microplastic images based on the Backbone network structure in the YOLO algorithm;

[0020] The Neck network structure is used to extract and fuse features of microplastic images, and the Predicton network structure is used to predict and output microplastic images.

[0021] Furthermore, constructing an automatic identification model for microplastics based on the YOLO algorithm includes at least the following steps:

[0022] Use labellmg to create labels for microplastic training images to form a json file, and then convert the json file to a txt file;

[0023] Based on the basic training model set above, the json file and the txt file are trained through train to form an automatic microplastic recognition model, and microplastic recognition is realized through detect.

[0024] Furthermore, the S3 at least includes the following steps:

[0025] Collect microplastics from dust samples and pre-treat them;

[0026] Then use high-definition video equipment to record images of microplastics;

[0027] The microplastic image is adjusted to the resolution of the input image to form a microplastic image to be processed for recognition by the microplastic automatic recognition model.

[0028] Furthermore, the S4 further includes perfecting the settings, and the perfecting the settings at least includes the following steps:

[0029] Analyzing the microplastics through an optical microscope to obtain optical microscope analysis results;

[0030] By comparing the optical microscope analysis results with the recognition results of the image to be processed, the reliability of the automatic recognition model is verified;

[0031] When the results are consistent, the reliability of the automatic recognition model is verified;

[0032] When the results are inconsistent, the inconsistent information in the automatic identification model will be improved and added to the database of typical characteristics of microplastics.

[0033] Furthermore, the recognition results of the image to be processed include picture results with confidence information, and EXCEL table results with the number of microplastic particles, microplastic morphological feature information, and microplastic particle color information.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] The present invention establishes an efficient, fast, easy-to-operate and low-cost intelligent quantification method for the number of microplastics based on machine vision learning, which greatly reduces the detection time. The types of microplastics can be distinguished by identifying the colors of the microplastics, which can solve the problem that existing methods are difficult to analyze microplastics in large quantities, and achieve accurate, efficient and economical intelligent batch identification of microplastics. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0037] Figure 1 This is a schematic diagram of the basic process of intelligent identification of microplastics based on machine vision learning in the present invention;

[0038] Figure 2 Schematic diagram of the process of the automatic identification method of microplastics of the present invention. DETAILED DESCRIPTION

[0039] 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, rather than all the embodiments.

[0040] See also Figure 1 - Figure 2 , the method for intelligently identifying the quantity and type of microplastics based on machine vision learning includes at least the following steps:

[0041] S1: Obtain the quantity and morphological characteristics of microplastics through experiments or other relevant methods, and combine and process them to form a database of typical characteristics of microplastics;

[0042] S2: Introduce machine learning and YOLO algorithm to build a basic recognition model, input the chemical property information and image information of microplastics into the basic image recognition model for targeted training, and form an automatic recognition model for microplastics. The basic model for automatic recognition of plastics can accurately identify the images in the database of typical features of microplastics, and has a certain recognition ability for new images;

[0043] S3: Acquire images of microplastics to be processed;

[0044] S4: Input the microplastic image to be processed into the automatic recognition model of microplastics, and obtain the recognition result of the image.

[0045] S2 inputs the chemical property information and image information of processed microplastics into the basic image recognition model for targeted training, which includes at least the following steps:

[0046] Classify the morphological features of microplastics in the training images to obtain multiple groups of microplastic training images;

[0047] Based on the morphological features of microplastics in the training images, the basic image recognition model is trained to form an automatic microplastic recognition system.

[0048] Based on the method of constructing a typical feature database of microplastics, the color and quantity of microplastics are associated with images in the form of electronic tags to form microplastic training images;

[0049] The microplastic training images were divided into training set, test set and validation set in a ratio of 5:4:1 for basic recognition model training.

[0050] Introducing machine learning and the YOLO algorithm to build a basic recognition model includes at least the following steps:

[0051] Preset the model selection, test data type, prediction result saving path, input image size, confidence threshold and result display settings in the YOLO algorithm to improve the basic model training speed and microplastic automatic identification model recognition speed;

[0052] Select yolov8s as the model of the YOLO algorithm, the test data type is image, the input image resolution is set to 640×640 according to the computer parameters, and the confidence threshold is set to 0.80 by default;

[0053] Cut microplastic images based on the Backbone network structure in the YOLO algorithm;

[0054] The Neck network structure is used to extract and fuse features of microplastic images, and the Predicton network structure is used to predict and output microplastic images.

[0055] When it is necessary to generate an automatic identification model, the number of training rounds is set to 200 based on the above basic training model, and the pre-set basic model is trained to ensure that the training results reach a sufficient confidence level, thereby forming a reliable microplastic automatic identification model.

[0056] Building an automatic identification model for microplastics based on the YOLO algorithm includes at least the following steps:

[0057] Use labellmg to create labels for microplastic training images to form a json file, and then convert the json file to a txt file;

[0058] Based on the basic training model set above, the json file and the txt file are trained through train to form an automatic microplastic recognition model, and microplastic recognition is realized through detect.

[0059] S3 includes at least the following steps:

[0060] Collect microplastics from dust samples and perform pre-treatment;

[0061] Then use high-definition video equipment to record images of microplastics;

[0062] The microplastic image is adjusted to the resolution of the input image to form a microplastic image to be processed for recognition by the microplastic automatic recognition model.

[0063] S4 also includes a complete setup, which includes at least the following steps:

[0064] Analyzing the microplastics through an optical microscope to obtain optical microscope analysis results;

[0065] By comparing the optical microscope analysis results with the recognition results of the image to be processed, the reliability of the automatic recognition model is verified;

[0066] When the results are consistent, the reliability of the automatic recognition model is verified;

[0067] When the results are inconsistent, the inconsistent information in the automatic identification model will be improved and added to the database of typical characteristics of microplastics.

[0068] The recognition results of the images to be processed include picture results with confidence information, EXCEL table results with the number of microplastic particles, microplastic morphological characteristics information, and microplastic particle color information.

[0069] Based on the above content, the working principle is summarized as follows:

[0070] First, the properties and image information of microplastics are obtained based on experiments or related methods, combined and processed to establish a database of typical characteristics of microplastics.

[0071] Secondly, by introducing machine learning and YOLO algorithm, a basic image recognition model is established.

[0072] After processing the chemical properties and image information of microplastics, they are input into the basic image recognition model for targeted training to form an automatic recognition model for microplastics.

[0073] Subsequently, the model is used to perform image recognition tasks and generate recognition results.

[0074] Finally, based on the basic model for automatic identification of microplastics, when the training results reach a sufficient confidence level, the image to be processed is obtained and input into the microplastic automatic identification model for recognition to obtain the recognition result of the image to be processed; at the same time, the recognition results are randomly selected for verification analysis to eliminate the influence of uncertain factors, improve the accuracy of the microplastic automatic identification results, verify the reliability of the model, and complete the iteration.

[0075] The identification results can determine the chemical property information, morphological feature information, and total number of microplastics in the image; the random sampling of identification results for verification and analysis is to use electron microscopy and Fourier transform infrared spectroscopy to conduct abundance and chemical property analysis on microplastic samples to verify and analyze the results.

[0076] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. An intelligent identification method for the quantity and type of microplastics based on machine vision learning, characterized by: At least the following steps are included: S1: Obtain the quantity and morphological characteristics of microplastics through experiments and scientific literature, combine and process them to form a database of typical characteristics of microplastics; S2: Introduce machine learning and YOLO algorithm to build a basic recognition model, input the chemical type and image information of microplastics into the basic image recognition model for targeted training, and form an automatic recognition model for microplastics. The basic model for automatic recognition of plastics can accurately identify the chemical type and particle quantity information of microplastics in the image in the microplastic typical feature database, and has a certain recognition ability for new images; S3: Acquire images of microplastics to be processed; S4: Input the microplastic image to be processed into the automatic recognition model of microplastics, and obtain the recognition result of the image.

2. The method for intelligently identifying the quantity and type of microplastics based on machine vision learning according to claim 1 is characterized in that: The step of inputting the chemical property information and image information of the processed microplastics into the basic image recognition model for targeted training in S2 includes at least the following steps: Classify the morphological features of microplastics in the training images to obtain multiple groups of microplastic training images; Based on the morphological features of microplastics in the training images, the basic image recognition model is trained to form an automatic microplastic recognition system. Based on the method of constructing a typical feature database of microplastics, the color and quantity of microplastics are associated with images in the form of electronic tags to form microplastic training images; The microplastic training images were divided into training set, test set and validation set in a ratio of 5:4:1 for basic recognition model training.

3. The method for intelligently identifying the quantity and type of microplastics based on machine vision learning according to claim 1 is characterized in that: The introduction of machine learning and YOLO algorithm to build a basic recognition model includes at least the following steps: The model selection, test data type, prediction result saving path, input image size, confidence threshold and result display settings in the preset YOLO algorithm are used to improve the training speed of the basic model and the recognition speed of the microplastic automatic recognition model; Select yolov8s as the model of the YOLO algorithm, set the test data type to image, set the input image resolution to 640×640 according to the computer parameters, and set the confidence threshold to 0.80 by default to complete the basic model settings; Cut microplastic images based on the Backbone network structure in the YOLO algorithm; The Neck network structure is used to extract and fuse features of microplastic images, and the Predicton network structure is used to predict and output microplastic images.

4. The method for intelligently identifying the quantity and type of microplastics based on machine vision learning according to claim 3 is characterized in that: Building an automatic identification model for microplastics based on the YOLO algorithm includes at least the following steps: Use labellmg to create labels for microplastic training images to form a json file, and then convert the json file to a txt file; Based on the basic training model set above, the json file and the txt file are trained through train to form an automatic microplastic recognition model, and microplastic recognition is realized through detect.

5. The method for intelligently identifying the quantity and type of microplastics based on machine vision learning according to claim 1 is characterized in that: The S3 at least includes the following steps: Collect microplastics from dust samples and pre-treat them; The microplastic images were then recorded using high-resolution optical microscopy and spectrometry; The microplastic image is adjusted to the resolution of the input image to form a microplastic image to be processed for recognition by the microplastic automatic recognition model.

6. The method for intelligently identifying the quantity and type of microplastics based on machine vision learning according to claim 1 is characterized in that: The S4 further includes perfecting the settings, and the perfecting the settings at least includes the following steps: Analyzing the microplastics through an optical microscope to obtain optical microscope analysis results; By comparing the optical microscope analysis results with the recognition results of the image to be processed, the reliability of the automatic recognition model is verified; When the results are consistent, the reliability of the automatic recognition model is verified; When the results are inconsistent, the inconsistent information in the automatic identification model will be improved and added to the database of typical characteristics of microplastics.

7. The method for intelligently identifying the quantity and type of microplastics based on machine vision learning according to claim 1, characterized in that: The recognition results of the image to be processed include picture results with confidence information, EXCEL table results with the number of microplastic particles, characteristic information of the microplastic types, and color information of the microplastic particles.

Images