【摘要】 塑料制品因其成本低廉且性能优良而广泛的应用于各行各业的方方面面，本质是高分子聚合物。但由于塑料在自然条件下降解困难且回收成本较高、回收效率较低，对环境造成了严重的影响。为解决塑料高效低成本的回收再利用问题，本文提出了一种基于近红外光谱技术，并结合当今前沿的机器学习技术的废杂塑料识别和分类系统。该系统采用基尼指数对近红外光谱进行特征波数选择，从而去除冗余数据。基于knn、LightGBM和神经网络建立单一的塑料分类模型，并采用随机搜索、遗传算法等算法对超参数进行优化。针对单一模型的不足之处，采用基于Stacking集成学习框架进行模型融合，进一步提高识别准确率和效率。结果表明该系统具有重要的实际运用价值和广阔的市场前景。更多还原

【Abstract】 Due to their low cost and excellent performance, the plastic products were widely used in all aspects of various industries, and the essence was polymer.However, due to the difficulty of desmobilizing plastics in natural conditions, the recovery cost was high, and the recycling efficiency was low, which had a serious impact on the environment.In order to solve the problem of high efficient and low-cost recycling of plastics, a waste plastic identification and classification system based on near-infrared spectroscopy technology combined with today’s cutting-edge machine learning technology was proposed.The system used the Gini index to select the characteristic wave number of the near-infrared spectra to remove redundant data.And a single plastic classification model based on knn, LightGBM and neural network was established, and used random search, genetic algorithm and other algorithms to optimize hyper parameters.In view of the shortcomings of a single model, the model fusion based on the Stacking integrated learning framework was adopted to further improve the recognition accuracy and efficiency.The results show that the system has important practical application value and broad market prospects.更多还原