Author Affiliations
11. School of New Energy, North China Electric Power University, Beijing 102206, China33. School of Artificial Intelligence, Jilin University, Changchun 130012, Chinashow less
Fig. 1. Original infrared spectra of cellulose (a), vinyl polymers (b), woods (c) and low-value wastes (d)
Fig. 2. SNV pretreated infrared spectra of cellulose (a), vinyl polymers (b), woods (c) and low-value wastes (d)
Fig. 3. MCS pretreated infrared spectra of cellulose (a), vinyl polymers (b), woods (c) and low-value wastes (d)
Fig. 4. DC/Smooth pretreated infrared spectra of cellulose (a), vinyl polymers (b), woods (c) and low-value wastes (d)
Fig. 5. The first (a) and second (b) principal component load analysis spectra
高值化类别 | 材料名称 |
---|
纤维素类 | 打印纸、 草纸、 一次性纸杯、 棉布、 烟头 | 烯类聚合物 | 方便面包装盒、 食品包装袋、 快餐包装纸、 奶茶杯、 腈纶标签 | 木竹类 | 竹扇、 落叶、 干树枝、 木质铅笔、 一次性筷子 | 低值类 | 棒骨、 陶瓷、 贝壳 |
|
Table 1. Residual waste materials
预处理方法 | 主成分 | Z1 | Z2 | Z3 | Z4 | Z5 | Z6 | Z7 | Z8 |
---|
| 特征值 | 208.35 | 102.04 | 50.50 | 39.22 | 32.71 | 14.99 | 6.91 | 4.15 | SNV | 方差贡献率/% | 43.7 | 21.4 | 10.6 | 8.2 | 7.1 | 3.1 | 1.5 | 0.9 | | 累计方差贡献率/% | 43.7 | 65.1 | 75.7 | 83.9 | 91.0 | 94.1 | 95.6 | 96.5 | | 特征值 | 861.19 | 429.45 | 207.22 | 161.67 | 107.20 | 47.83 | 36.77 | 24.15 | MSC | 方差贡献率/% | 44.6 | 22.2 | 10.7 | 8.4 | 5.6 | 2.5 | 1.9 | 1.2 | | 累计方差贡献率/% | 44.6 | 66.8 | 77.5 | 85.9 | 91.5 | 94.0 | 95.9 | 97.1 | | 特征值 | 149.73 | 80.39 | 16.76 | 14.34 | 6.66 | 4.98 | 3.28 | 2.50 | DC/Smooth | 方差贡献率/% | 52.2 | 28.0 | 5.8 | 5.0 | 2.3 | 1.8 | 1.1 | 0.9 | | 累计方差贡献率/% | 52.2 | 80.2 | 86.0 | 91.0 | 93.3 | 95.1 | 96.2 | 97.1 |
|
Table 2. The principal component eigenvalues and variance contribution of SNV, MSC and DC/smooth datasets
预处理方式 | 分类准确率/% | 均值 /% | 均方根 误差/% |
---|
PNN | GRNN | RDF | SVM |
---|
未预处理 | 78.7 | 85.7 | 85.7 | 81.5 | 82.9 | 3.4 | SNV | 88.7 | 90.1 | 88.7 | 90.1 | 89.4 | 0.8 | MSC | 87.3 | 90.1 | 87.3 | 88.7 | 88.4 | 1.3 | DC/Smooth | 95.8 | 87.5 | 95.8 | 97.2 | 94.1 | 4.4 | 均值* | 90.6 | 89.2 | 90.6 | 92.0 | - | - | 均方根误差* | 4.6 | 1.5 | 4.6 | 4.6 | - | - |
|
Table 3. Comparison of classification model accuracies (based on 72×8 dataset)
预处理方式 | 分类准确率/% | 均值 /% | 均方根 误差/% |
---|
PNN | GRNN | RDF | SVM |
---|
未预处理 | 77.8 | 84.7 | 84.7 | 87.5 | 83.6 | 4.1 | SNV | 95.8 | 100.0 | 88.9 | 90.3 | 93.8 | 5.1 | MSC | 98.6 | 88.9 | 90.2 | 91.7 | 92.4 | 4.3 | DC/Smooth | 100.0 | 94.4 | 95.8 | 95.8 | 96.5 | 2.4 | 均值* | 98.1 | 94.4 | 91.6 | 92.6 | - | - | 均方根误差* | 2.1 | 5.6 | 3.6 | 2.8 | - | - |
|
Table 4. Comparison of classification model accuracies (based on 72×5 dataset)
垃圾类别 | DC/Smooth | 均值 /% | 均方根 误差/% |
---|
PNN | GRNN | RDF | SVM |
---|
纤维素类 | 95.0 | 90.0 | 100.0 | 95.0 | 95.0 | 4.1 | 烯类聚合物 | 95.0 | 80.0 | 90.0 | 95.0 | 90.0 | 7.1 | 木竹类 | 95.0 | 90.0 | 95.0 | 100.0 | 95.0 | 4.1 | 低值类 | 100.0 | 91.6 | 100.0 | 100.0 | 97.9 | 4.2 |
|
Table 5. Comparison of classification accuracies for the four kinds of residual wastes (based on 72×8 DC/Smooth dataset)
垃圾类别 | DC/Smooth | 均值 /% | 均方根 误差/% |
---|
PNN | GRNN | RDF | SVM |
---|
纤维素类 | 100.0 | 100.0 | 95.0 | 95.0 | 97.5 | 2.9 | 烯类聚合物 | 100.0 | 90.0 | 95.0 | 90.0 | 93.8 | 4.8 | 木竹类 | 100.0 | 90.0 | 95.0 | 100.0 | 96.3 | 4.8 | 低值类 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 0.0 |
|
Table 6. Comparison of classification accuracies for the four kinds of residual wastes (based on 72×5 DC/Smooth dataset)