Prediction Of Calorific Value Of Domestic Waste Based On Near-infrared Hyperspectral

With the rapid development of economy,non-renewable resources are exploited in large quantities,and the development of new energy sources is advocated globally.In the fast-paced life,people generate a large amount of domestic waste every day.How to maximize the utilization of domestic waste more efficiently and reduce the pollution to the environment? The subject comes from the Chongqing Institute of Waste Incineration Power Generation Technology.According to the needs of industrial incineration power generation,the intelligent prediction of the calorific value of domestic waste is realized.In order to more accurately calculate the calorific value of combustion of various waste components,the hyperspectral images of domestic waste were first collected to make a data set,then the principal components were classified,and finally the overall calorific value of combustion was calculated according to the classification results.Due to the problem that ordinary cameras are difficult to image clearly in the case of weak textures,and the existing thermal value estimation techniques and algorithms are applied in complex industrialized environments,there are problems such as poor stability,high cost,and poor real-time performance.Therefore,this paper proposes a method for predicting the calorific value of domestic waste based on near-infrared hyperspectral.The main research contents are as follows:1.In view of the high cost of industrial applications,it is difficult for ordinary cameras to clearly image under weak textures,and the same type of substances are affected by visible light and produce different spectral data.A near-infrared hyperspectral camera combined with a short-wave centroid camera and an LCTF filter,and a hyperspectral image data acquisition system is built to complete the collection of five common household waste samples of wood,paper,plastic,textiles,and foam.The 920～1700nm band range Experiments were performed on the hyperspectral image data within.The experimental results show that the self-built nearinfrared hyperspectral camera reduces the cost of industrialization,and the self-collected hyperspectral image data is consistent with the standard experimental data.2.Aiming at the problems of over-fitting and poor real-time performance of the model caused by the excessive amount of hyperspectral data.Feature band extraction is performed on the collected raw hyperspectral data to reduce the amount of redundant information in hyperspectral image data.In order to extract more accurate feature band data,the 3D image data is preprocessed first,and then the contribution rate-based and weight-based methods are used.eigenband algorithm.The data set after the extraction of bands is used to replace the original data set for classification research,and the SVM classification algorithm is used to verify the extracted characteristic bands.The experimental results show that the feature bands extracted by this method have good intra-class similarity and inter-class difference.3.In view of the Hughes phenomenon caused by the deepening of the network layer and the insufficient utilization of spatial spectral feature information,a residual module of branch structure is proposed to extract rich spatial spectral features by 3D convolution of different scales and sizes of two-dimensional images and spectral information.First,the performance of the model is verified by the public dataset IP,and then the dataset after band processing is used for classification.The experimental results show that the classification accuracy of the 3DMRCNN model reaches 96%,which has better accuracy and real-time performance than the traditional classification model and the MSDN model.4.Based on the above algorithm,linearly calculate the estimated calorific value of the classification result and the unit calorific value of the principal components of various types of domestic waste.The experimental results show that the estimation error of this method is small,and it is feasible to apply it to intelligent calorific value estimation of industrial incineration.