主成分分析结合支持向量机辅助激光诱导击穿光谱对塑料快速分类识别

大量废弃的塑料制品给生态环境造成严重破坏，当务之急是要对塑料进行分类回收。传统的分类方法普遍存在成本高，效率低，操作复杂等问题，不能满足工业生产的需要。激光诱导击穿光谱技术由于具备简单灵活，快速灵敏等优点，在物质鉴别领域有广泛应用。采用激光诱导击穿光谱技术结合主成分分析（PCA）和支持向量机（SVM）算法对20种塑料进行分类识别研究。由于目前有关塑料分类识别速率的研究报道较少，该实验在保证识别准确率的前提下，进一步研究和分析实验过程耗费的时间，满足工业生产中快速分类的要求。每种塑料采集100组光谱数据，随机选取50组数据作为训练集建立模型，余下50组作为测试集测试模型的分类识别效果，所以训练集和测试集各有1 000组光谱数据。将训练集的数据不加处理地输入SVM中进行训练，并采用5折交叉验证建立最佳模型，此时测试集的识别准确率为99.90%，建模时间为1小时58分41.13秒，预测时间为11.96 s。由此可见，单纯使用SVM算法可以得到很高的准确率，但是需要耗费大量时间。为了提高实验效率，引入主成分分析算法，将原来的高维数据变换成低维数据，并用降维后的数据训练模型。针对不同的主成分个数，均采用随机训练十次再取平均值的方法获得相关数值。实验表明，当选取主成分个数为13时，得到相应的识别准确率为99.80%，而PCA处理时间为1.44 s，建模时间为12.16 s，预测时间仅为0.02 s。虽然PCA算法结合SVM算法在对20种塑料进行分类识别时准确率有轻微下降，但是大大减少了模型训练的时间，实验效率得到很大程度的提高。结果表明，结合两种算法辅助激光诱导击穿光谱可以对塑料进行快速准确的分类识别。

Rapid Classification and Identification of Plastic Using Laser-Induced Breakdown Spectroscopy With Principal Component Analysis and Support Vector Machine

A large number of discarded plastic products cause serious damage to the ecological environment. It is urgent to recycle plastic by classification. The traditional classification method can not meet the needs of industrial production due to its high cost, low efficiency and complex operation. Laser-induced breakdown spectroscopy (LIBS) has been widely used in the field of substance identification with many advantages, such as simplicity, flexibility, speed and sensitivity. In this paper, 20 kinds of plastics were classified and identified by LIBS combined with principal component analysis (PCA) and support vector machine (SVM). Since few papers have studied the classification and recognition rate of plastic at present, the experiment further studies and analyzes the time spent in the experimental process on the premise of ensuring the accuracy of identification, so as to meet the requirements of rapid classification in industrial production. During the study, 100 groups of spectral data were collected for each plastic, 50 groups of data were randomly selected as the training set to establish the model, and the remaining 50 groups were used as a test set to validate model. Therefore, the training set and the test set each had 1 000 groups of spectral data. The data of the training set was input into SVM for training without any processing, and the best model was established by using the five-fold cross validation. At this time, the recognition accuracy of the test set was 99.90%, the modeling time was 1 hour, 58 minutes, 41.13 seconds, and the prediction time was 11.96 seconds. Thus, it can be seen that the SVM algorithm can be used simply to achieve high accuracy, but it needs a lot of time. In order to improve the experimental efficiency, a principal component analysis algorithm is introduced to process the data, transform the original high-dimensional data into low-dimensional data, and train the model with the data after dimension reduction. For different principal component numbers, the experimental values were obtained by random training ten times and taking the mean value. Experiments show that when the number of principal components is 13, the corresponding recognition accuracy is 99.80%, while PCA processing time is 1.44 seconds, modeling time is 12.16 seconds, and prediction time is only 0.02 seconds. Although the PCA algorithm combined with the SVM algorithm has a slight decrease in the accuracy of classification and recognition for 20 kinds of plastics, it greatly reduces the time of model training and greatly improves the experimental efficiency. The results show that the two algorithms can be used to classify and identify plastic quickly and accurately.