Plastic material identification method based on near infrared spectrum wavelength ratio reconstruction characteristics

Abstract

The invention discloses a plastic material identification method based on near infrared spectrum wavelength ratio reconstruction characteristics, and belongs to the technical field of spectrum identification, and the method comprises the steps: 1, selecting a specific number of wavelengths of effective wave bands according to the specific type of plastic during identification, so as to obtain the spectral reflectivity value of each effective wave band; and 2, constructing spectral characteristics according to the spectral reflectivity value of each effective wave band, and identifying the plastic based on the spectral characteristics. According to the method, the corresponding spectral characteristics can be constructed according to the spectral reflectivity conditions corresponding to the dual wavelengths and the four wavelengths, so that the plastics are identified, accurate identification can be performed on one or two specific types of plastics, and the plastic identification efficiency can be greatly improved under the scene that multiple types of plastics need to be sorted out at the same time. And various plastics are accurately identified, so that the optimal identification effect of the plastics is realized under the condition of limited cost.

Description

Technical Field

[0001] The present invention belongs to the technical field of spectrum recognition, and in particular relates to a plastic material recognition method based on near-infrared spectrum wavelength ratio reconstruction features. Background Art

[0002] With the widespread use of plastic products worldwide, the disposal and recycling of plastic waste has become a pressing global issue. Accurate identification of plastic materials, a key step in achieving efficient sorting and recycling, directly impacts resource recycling efficiency and environmental protection. Furthermore, in areas such as product quality testing and medical packaging safety monitoring, precise identification of plastic materials is a crucial prerequisite for ensuring product performance and safety.

[0003] Traditional near-infrared spectroscopy plastic material identification methods often employ spectrometers and use multiple characteristic wavelengths for analysis and identification. This increases identification costs and data processing complexity, hindering industrial application. Furthermore, in actual plastic recycling applications, there is sometimes a need for efficient sorting of only one or two specific types of plastic. For example, for PET plastics, using multiple characteristic wavelengths for analysis and identification increases computational cost and time. Sometimes, there is a need to simultaneously sort multiple types of plastic. While using multiple characteristic wavelengths for analysis and identification can improve identification accuracy, this significantly increases identification costs, hindering industrial application. Summary of the Invention

[0004] The purpose of the present invention is to provide a plastic material identification method based on near-infrared spectrum wavelength ratio reconstruction features to solve the problems faced in the above-mentioned background technology.

[0005] The purpose of the present invention can be achieved through the following technical solutions: A plastic material recognition method based on near-infrared spectrum wavelength ratio reconstruction features, the method comprising: Step 1: Select a specific number of effective wavelength bands according to the specific type of plastic being identified, thereby obtaining the spectral reflectance value of each effective wavelength band; Step 2: Construct spectral features based on the spectral reflectance values of each effective band, and identify plastics based on the spectral features; The method for constructing spectral features in step 2 is: select two sets of effective bands at wavelengths of 1120nm and 1190nm, and obtain the spectral reflectance values of each band respectively 、 , let the original ratio For the spectral intensity ratio feature, the spectral intensity ratio feature is optimized by using the nonlinear expansion difference optimization method. The optimization method includes: inverse proportional function Optimization, tangent function Optimization and ratio feature normalization optimization.

[0006] Furthermore, the inverse proportional function The optimization method is: Based on the inverse proportional function characteristics, all the original ratio data are shifted to the left by 0.95, that is, the inverse proportional value of the ratio shift is , thereby optimizing the spectral intensity ratio characteristics, where ; Then Compare with the preset threshold intervals to identify the plastic material.

[0007] Furthermore, the tangent function The optimization method is: Based on the characteristics of the tangent function, all the original ratio data are right-shifted by 0.55, that is, the ratio shift tangent value is , thereby optimizing the spectral intensity ratio characteristics, where ; Then Compare with the preset threshold intervals to identify the plastic material.

[0008] Furthermore, the ratio feature normalization optimization method is: First, normalize the original ratio and select the normalized interval range ; Then find the minimum value of the original ratio data from the historical database and the maximum value , through the formula Calculate the coefficient value ; So as to normalize to Data value of the interval , ; Finally, the data value is compared with the preset threshold intervals to identify the plastic material.

[0009] Furthermore, the method for constructing spectral features in step 2 further includes: Select four groups of effective wavelengths at 1120nm, 1190nm, 1639nm, and 1716nm, and obtain the corresponding spectral reflectance at each group of wavelengths 、 、 、 , using the spectral reflectance ratio of the first two wavelengths The spectral reflectance ratio corresponding to the last two wavelengths , constructing four-wavelength spectral features , , , based on the four-wavelength spectral characteristics Build a plastic recognition model.

[0010] Furthermore, based on the four-wavelength spectral characteristics The method of constructing the plastic recognition model is as follows: using the four-wavelength spectral feature W and constructing the recognition model based on the KNN algorithm; According to the ratio of 7:3, the four-wavelength spectral features W are randomly divided into a training set and a test set. The training set is used for model construction, the training set is used for model training, and the test set is used for accuracy evaluation.

[0011] Beneficial effects of the present invention: The present invention constructs a dual-wavelength spectral intensity ratio feature by identifying the spectral reflectance values of the plastic's effective band at 1120nm and 1190nm, and optimizes the feature accordingly, so as to accurately identify the plastic material based on the optimized value. In this way, when only one or two specific types of plastics need to be identified, more focus can be placed on the unique spectral features of the plastic, thereby realizing the characteristic identification of the target plastic, avoiding the unnecessary step of searching for non-target plastic features in complex spectral data, improving the accuracy and efficiency of identification, and reducing computing cost and time.

[0012] The present invention constructs a four-wavelength spectral feature by obtaining the spectral reflectance at effective bands of 1120nm, 1190nm, 1639nm, and 1716nm, and constructs an identification model and evaluates the accuracy based on the constructed spectral feature. This method can improve the comprehensiveness and accuracy of identification by adding characteristic wavelengths in scenarios where multiple types of plastics need to be sorted out simultaneously, enabling it to cope with more diverse plastic samples and complex recycling environments, and also ensure the best identification effect at a limited cost.

[0013] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] 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.

[0015] Figure 1 is a flow chart of the method of the present invention; Figure 2 is a graph showing the average difference in the spectral intensity ratios of the various plastics in the present invention; Figure 3 It is the average difference diagram of the inverse proportional value of the spectral shift of each plastic in the present invention; Figure 4 is the average difference diagram of the tangent shift values of the spectra of each plastic in the present invention; Figure 5 It is the average difference graph of the normalized data values of each plastic spectral data in the present invention; Figure 6 A difference diagram of the four wavelength spectral features selected from the near-infrared average spectrum of each plastic in the present invention; Figure 7 The spectral characteristics of the present invention Axis scatter plot of ; Figure 8 This is a graph representing the accuracy of the classification of various plastics in the present invention. DETAILED DESCRIPTION

[0016] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0017] In one embodiment, a plastic material identification method based on near-infrared spectrum wavelength ratio reconstruction features is disclosed, such as Figure 1 As shown, the identification method mainly includes: Step 1: Select a specific number of effective wavelength bands according to the specific type of plastic being identified, thereby obtaining the spectral reflectance value of each effective wavelength band; Step 2: Construct spectral features based on the spectral reflectance values of each effective band, and identify plastics based on the spectral features.

[0018] Through the above technical solution, the present application obtains the spectral reflectance values of the effective band for identifying plastics at 1120nm and 1190nm to construct a dual-wavelength spectral intensity ratio feature, and optimizes the feature accordingly, so as to accurately identify the material of the plastic according to the optimized value. In this way, when only one or two specific types of plastics need to be identified, the identification method of a small number of wavelengths can focus more on the unique spectral characteristics of the plastic, thereby realizing the characteristic identification of the target plastic, avoiding the unnecessary step of searching for non-target plastic characteristics in complex spectral data, and improving the accuracy and efficiency of identification; at the same time, using a small number of wavelengths for identification and analysis can simplify the data processing process, reduce computing costs and time, and through the mechanical energy identification of a small number of carefully selected wavelengths, the response speed of the system can be significantly improved, making the entire sorting process smoother and more efficient. By obtaining the spectral reflectance of the effective bands at 1120nm, 1190nm, 1639nm, and 1716nm, a four-wavelength spectral feature is constructed. The recognition model is constructed and the accuracy is evaluated based on the constructed spectral feature. This method can improve the comprehensiveness and accuracy of recognition by increasing the characteristic wavelengths when there is a need to sort out multiple types of plastics at the same time. Moreover, the reasonable increase of wavelengths (a total of four groups) not only helps to build more features and improve the generalization ability so that it can cope with more diverse plastic samples and complex recycling environments, but also ensures the best recognition effect at a limited cost.

[0019] The method for constructing spectral features in step 2 is: select two sets of effective bands at wavelengths of 1120nm and 1190nm, and obtain the spectral reflectance values of each band respectively 、 , let the original ratio For the spectral intensity ratio feature, the spectral intensity ratio feature is optimized by using the nonlinear expansion difference optimization method. The optimization method includes: inverse proportional function Optimization, tangent function Optimization and ratio feature normalization optimization.

[0020] The above scheme provides a specific method for constructing spectral intensity ratio characteristics using two wavelengths. In the actual application scenario of plastic recycling, there is a need for efficient sorting of only one or two specific types of plastics, such as PET plastics. The identification method based on a small number of wavelengths can focus more on the unique spectral characteristics of the plastic, thereby realizing the characteristic identification of the target plastic. Therefore, in the identification and sorting of PET, a large number of experiments have shown that the effective wavelength band of PET is around 1120nm and 1190nm. Therefore, the spectral reflectance value at 1120nm is set to , the spectral reflectance value at 1190nm is , so through the formula The spectral intensity ratio characteristics are obtained and identified based on the original ratio K. In this experiment, seven common plastics (PET, PE, HDPE, PP, PS, PVC, ABS) were tested to study the numerical differences in their ratios at specific wavelengths. 45 samples of each plastic material were selected, including samples of different colors and thicknesses. The experimental results are as follows: Figure 2 As shown in the experiment, it was found that the difference in K values was not obvious, and the various types of plastics were not distinguishable. The numerical differences were relatively small, so optimization was needed. The optimization method of nonlinear expansion of differences could be used, including inverse proportional function. Optimization, tangent function Optimization and ratio feature normalization optimization, etc.

[0021] Inverse proportional function The optimization method is: based on the inverse proportional function characteristics, all the original ratio data are shifted to the left by 0.95, that is, the inverse proportional value of the ratio shift is , thereby optimizing the spectral intensity ratio characteristics, where ; Then Compare with the preset threshold intervals to identify the plastic material.

[0022] Due to the inverse proportional function, hour, , , the greater the difference in the value of f(x), the pre-experiment found that the original ratios are all greater than 0.98 and above, so the closer the left translation value is to 0, the greater the difference, so all the original ratio data are left-translated by 0.95 to obtain the inverse proportional value of the translation , each plastic spectrum shifts inversely proportional values The average difference of Figure 3 As shown, it can be concluded that after translation and then inversely proportional amplification of the difference, If the value is greater than 15, it is PET plastic, otherwise it is other plastics. This method can set a threshold value to effectively recycle PET plastics with large recycling volume and high economic benefits.

[0023] Tangent function The optimization method is: based on the characteristics of the tangent function, all the original ratio data are right-shifted by 0.55, that is, the ratio shift tangent value is , thereby optimizing the spectral intensity ratio characteristics, where ; Then Compare with the preset threshold intervals to identify the plastic material.

[0024] Since the tangent function is hour, , ; , Based on this characteristic, the feature difference of the ratio is amplified. Preliminary experiments show that the closer the original ratio is to the right, the The effect is better, so the value of right shift is closer to , the greater the difference, so choose right translation 0.55. That is, let the ratio translation tangent value be , , get the tangent value of each plastic spectrum The average difference of Figure 4 As shown, after translation and tangent amplification, for If the threshold is greater than 0, it is PET plastic, otherwise it is other plastics. This method can set the threshold to 0, and PET plastic can be effectively recycled. At the same time, for PS plastic, when the threshold is less than -70, effective features can also be highlighted, thereby separating PET and PS. In order to verify the effect of threshold selection in plastic type identification, 700 samples were used for verification, and the samples were subjected to threshold identification based on the difference characteristics of the spectral shift tangent function. The experimental results show that when When it is greater than 0, all samples are accurately identified as PET plastics, and no other types of plastics are misidentified as PET. When the value is less than -70, the sample is classified as PS plastic. Under this method, both PET and PS can be selected.

[0025] The optimization method for the ratio feature normalization in step 2 is: first normalize the original ratio and select the normalization interval range ; Then find the minimum value of the original ratio data from the historical database and the maximum value , through the formula Calculate the coefficient value ; So as to normalize to Data value of the interval , ; Finally, the data value is compared with the preset threshold intervals to identify the plastic material.

[0026] Since normalization can convert data with different features to the same scale or range, it can ensure that all features have similar weights, thereby improving accuracy and stability. In addition, after normalization, it can also be seen whether there are differences in the features of some data. To normalize the ratio, you can choose to normalize the ratio to [-1, 1], and then find the minimum value of the original ratio data from the historical database. and the maximum value , through the formula Calculate the coefficient value ; thus normalized to Data value of the interval , ; Preliminary experiments found that the maximum value of the original ratio was not greater than 1.5 and the minimum value was not less than 0.9. Therefore, the maximum value of 1.5 and the minimum value of 0.9 were selected from the obtained spectral intensity ratio data and substituted into the formula The average difference of the normalized data values of each plastic spectrum data is obtained as shown in the figure Figure 5 As shown in the figure, after normalizing the original ratio data of plastic, it can be seen that When the threshold is less than -0.6, it is PET plastic. At the same time, it can be observed that for PP plastic, when the threshold is greater than 0, it is PP plastic, which effectively separates PET and PP. In order to verify the effect of threshold selection in plastic type identification, 700 samples were used for verification and the samples were subjected to spectral normalization function difference feature threshold identification. The experimental results show that when When the value is less than the threshold value -0.6, all samples are accurately identified as PET plastics, and no other types of plastics are found mixed in them. When the value is greater than the threshold of 0, the sample is classified as PP plastic. Under this method, both PET and PS can be selected.

[0027] The method for constructing spectral characteristics in step 2 also includes: selecting four groups of effective wavelengths at 1120nm, 1190nm, 1639nm, and 1716nm, and obtaining the corresponding spectral reflectance at each group of wavelengths 、 、 、 , using the spectral reflectance ratio of the first two wavelengths The spectral reflectance ratio corresponding to the last two wavelengths , constructing four-wavelength spectral features , , , based on the four-wavelength spectral characteristics Constructing a plastic recognition model based on four-wavelength spectral characteristics The method for constructing a plastic recognition model is as follows: using the four-wavelength spectral feature W, the recognition model is constructed based on the KNN algorithm, and the four-wavelength spectral feature W is randomly divided into a training set and a test set in a ratio of 7:3. The training set is used for model construction, the model is trained with the training set, and the accuracy is evaluated with the test set.

[0028] The above solution provides a specific method for constructing spectral signatures using four wavelengths. In actual plastic recycling operations, multiple types of plastics must be sorted simultaneously to meet different recycling processes or market demands. As the number of plastic types to be identified increases, the unique spectral signatures of each plastic may overlap or become similar, increasing the difficulty of identification. In this case, adding more characteristic wavelengths is necessary to improve the comprehensiveness and accuracy of identification. By adding more characteristic wavelengths, more subtle spectral differences can be captured, enabling more accurate differentiation between different plastics. Adding more characteristic wavelengths helps construct more features and improves generalization, enabling it to handle a wider variety of plastic samples and complex recycling environments. While increasing characteristic wavelengths to improve identification accuracy, cost-effectiveness must also be considered. Therefore, the number of wavelengths selected should be limited to ensure optimal identification performance within a limited cost. Experiments were conducted on seven common plastic materials (PET, PE, HDPE, PP, PS, PVC, and ABS). Forty-five samples of each plastic material were selected, encompassing samples of varying colors and thicknesses. A plastic's characteristic wavelength primarily reflects its absorption and reflection properties at different wavelengths of light. These properties are closely related to factors such as the plastic's type, structure, and additives. The study found that the characteristics of waste plastics in the near-infrared spectroscopy are reflected in some main wavelengths. Extracting and constructing new features at these wavelengths is conducive to plastic identification and sorting. These characteristic wavelengths may include but are not limited to 1120nm, 1190nm, 1216nm, 1220nm, 1394nm, 1401nm, 1639nm, 1660nm, 1716nm, etc. The experiment selects wavelengths with the goal of distinguishing multiple types of plastics. After a large number of experiments and calculations, it was found that the reflectivity of the four wavelengths of 1120nm, 1190nm, 1639nm, and 1716nm is the best, so four groups of wavelengths are selected: 1120nm, 1190nm, 1639nm, and 1716nm, and the corresponding reflectivity is selected at each group of wavelengths. 、 、 、 , using the reflectivity ratio of the first two wavelengths And the reflectivity ratio corresponding to the last two wavelengths , construct spectral features , , The spectral characteristics of the four wavelengths of the seven plastics are as follows: Figure 6 As shown in the figure, it can be seen that the two-segment ratio characteristics of different materials are different. 、 The two features distinguish these seven types of plastics. PET, PP, PS, and ABS can be distinguished, but there is a certain degree of overlap between the ratio features of PE, HDPE, and PVC plastics, resulting in three or more materials being unable to be distinguished. Therefore, the ratio feature W is selected as the plastic material identification feature to facilitate the classification results. The spectral feature The axis scatter plot is as follows Figure 7 As shown, the ratio feature W is input into the KNN algorithm to build a plastic material recognition model. The training set and test set are randomly divided into training set and test set with a ratio of 7:3, and the test set accuracy is used for evaluation. It is found that PET, PP, PS, and ABS are 100% recognized from the seven types of plastics, while the recognition accuracy of more than five categories is as follows: Figure 8 As shown, this method uses only four wavelengths to construct signatures, resulting in a relatively simple identification feature. This not only simplifies the feature composition and algorithm complexity, but also reduces the requirements for spectral measurement. This provides a technical reference for the development of industrial application systems that can efficiently and cost-effectively identify recycled plastic materials.

[0029] The above content is merely an example and explanation of the concept of the present invention. Those skilled in the art may make various modifications or additions to the described specific embodiments or replace them in a similar manner. As long as they do not deviate from the concept of the invention or exceed the scope defined by the claims, they should all fall within the scope of protection of the present invention.

Claims

1. A plastic material identification method based on near-infrared spectrum wavelength ratio reconstruction features, characterized in that: The method comprises: Step 1: Select a specific number of effective wavelength bands according to the specific type of plastic being identified, thereby obtaining the spectral reflectance value of each effective wavelength band; Step 2: Construct spectral features based on the spectral reflectance values of each effective band, and identify plastics based on the spectral features; The method for constructing spectral features in step 2 is: select two sets of effective bands at wavelengths of 1120nm and 1190nm, and obtain the spectral reflectance values of each band respectively 、 , let the original ratio For the spectral intensity ratio feature, the spectral intensity ratio feature is optimized by using the nonlinear expansion difference optimization method. The optimization method includes: inverse proportional function Optimization, tangent function Optimization and ratio feature normalization optimization.

2. The plastic material identification method based on near-infrared spectrum wavelength ratio reconstruction feature according to claim 1 is characterized in that: The inverse proportional function The optimization method is: Based on the inverse proportional function characteristics, all the original ratio data are shifted to the left by 0.95, that is, the inverse proportional value of the ratio shift is , thereby optimizing the spectral intensity ratio characteristics, where ; Then Compare with the preset threshold intervals to identify the plastic material.

3. The plastic material identification method based on near-infrared spectrum wavelength ratio reconstruction feature according to claim 1 is characterized in that: The tangent function The optimization method is: Based on the characteristics of the tangent function, all the original ratio data are right-shifted by 0.55, that is, the ratio shift tangent value is , thereby optimizing the spectral intensity ratio characteristics, where ; Then Compare with the preset threshold intervals to identify the plastic material.

4. The plastic material identification method based on near-infrared spectrum wavelength and ratio reconstruction features according to claim 1 is characterized in that: The ratio feature normalization optimization method is: First, normalize the original ratio and select the normalized interval range ; Then find the minimum value of the original ratio data from the historical database and the maximum value , through the formula Calculate the coefficient value ; So as to normalize to Data values of the interval , ; Finally, the data value is compared with the preset threshold intervals to identify the plastic material.

5. The plastic material identification method based on near-infrared spectrum wavelength ratio reconstruction feature according to claim 1 is characterized in that: The method for constructing spectral features in step 2 further includes: Select four groups of effective wavelengths at 1120nm, 1190nm, 1639nm, and 1716nm, and obtain the corresponding spectral reflectance at each group of wavelengths 、 、 、 , using the spectral reflectance ratio of the first two wavelengths The spectral reflectance ratio corresponding to the last two wavelengths , constructing four-wavelength spectral features , , , based on the four-wavelength spectral characteristics Build a plastic recognition model.

6. The plastic material identification method based on near-infrared spectrum wavelength ratio reconstruction feature according to claim 1 is characterized in that: Based on four-wavelength spectral characteristics The method of constructing the plastic recognition model is as follows: using the four-wavelength spectral feature W and constructing the recognition model based on the KNN algorithm; According to the ratio of 7:3, the four-wavelength spectral features W are randomly divided into a training set and a test set. The training set is used for model construction, the training set is used for model training, and the test set is used for accuracy evaluation.

Images