Low-Cost Recognition of Plastic Waste Using Deep Learning and a Multi-Spectral Near-Infrared Sensor

Abstract

This work presents an approach for the recognition of plastics using a low-cost spectroscopy sensor module together with a set of machine learning methods. The sensor is a multi-spectral module capable of measuring 18 wavelengths from the visible to the near-infrared. Data processing and analysis are performed using a set of ten machine learning methods (Random Forest, Support Vector Machines, Multi-Layer Perceptron, Convolutional Neural Networks, Decision Trees, Logistic Regression, Naive Bayes, k-Nearest Neighbour, AdaBoost, Linear Discriminant Analysis). An experimental setup is designed for systematic data collection from six plastic types including PET, HDPE, PVC, LDPE, PP and PS household waste. The set of computational methods is implemented in a generalised pipeline for the validation of the proposed approach for the recognition of plastics. The results show that Convolutional Neural Networks and Multi-Layer Perceptron can recognise plastics with a mean accuracy of 72.50% and 70.25%, respectively, with the largest accuracy of 83.5% for PS plastic and the smallest accuracy of 66% for PET plastic. The results demonstrate that this low-cost near-infrared sensor with machine learning methods can recognise plastics effectively, making it an affordable and portable approach that contributes to the development of sustainable systems with potential for applications in other fields such as agriculture, e-waste recycling, healthcare and manufacturing.

Keywords: low-cost sensors; machine learning; near-infrared sensor; plastic recognition; principal component analysis.

Figure 1

Low-cost spectroscopy sensor and examples of plastic waste. (a) Triad spectral sensor module from SparkFun Electronics [41]. (b) Examples of household plastic waste used for data collection and recognition processes.

Figure 2

Illustration of the experimental setup used for systematic data collection. (a) View of the setup with the sensor and a plastic samples. (b) Setup with physical dimensions and illustration of the sensor measurement volume.

Figure 3

Example of data collected from plastic waste and visualisation with dimensionality reduction. (a–f) Spectral information with 200 measurements from each of the six plastic types. Two main components from the plastic waste using (g) Principal Component Analysis and (h) Linear Discriminant Analysis techniques.

Figure 3

Example of data collected from plastic waste and visualisation with dimensionality reduction. (a–f) Spectral information with 200 measurements from each of the six plastic types. Two main components from the plastic waste using (g) Principal Component Analysis and (h) Linear Discriminant Analysis techniques.

Figure 4

Generalised approach for implementation, training and testing of the set of machine learning methods composed of cross-validation, dimensionality reduction, recognition, performance metrics and optimisation stages.

Figure 5

Meandata from plastic waste samples collected in Section 3.4.

Figure 6

Results from plastic waste and each computational method. (a) Mean recognition accuracy from each machine learning method using (i) raw data only, (ii) PCA preprocessing and (iii) LDA preprocessing. (b–d) Confusion matrices with the highest recognition accuracy for each plastic type using machine learning and raw data, LDA and PCA.