Author Affiliations
11. School of Mechatronics and Vehicle Engineering, East China Jiaotong University, Nanchang 330013, China22. Technology Center of Nanchang Customs District, Nanchang 330038, Chinashow less
Fig. 1. Schematic diagram of the vis-near infrared spectroscopy online sorting device for ‘Yali' pear
Fig. 2. Arrangement top view of halogen lamp
Fig. 3. Energy spectra curve of normal pear and black heart pear
Fig. 4. Distribution of the first three principal components of normal pears and black heart pears
Fig. 5. AdaBoost algorithm principle
Fig. 6. Comparison of actual categories and predicted categories in WT-AdaBoost model for ‘Yali' pear samples
Table 1. Sample set information
真实情况 | 预测结果 |
---|
正常梨 | 黑心梨 |
---|
正常梨 | TP | FN | 黑心梨 | FP | TN |
|
Table 2. confusion matrix for classification result
预处理方法 | F-measure/% | Accuracy/% |
---|
Raw | 77.38 | 80.00 | Smooth | 73.49 | 76.84 | SNV | 77.38 | 80.00 | MSC | 77.38 | 80.00 | SG 1st-Der | 72.11 | 78.42 | WT | 78.98 | 82.62 |
|
Table 3. kNN model results of qualitative identification of ‘Yali' pears with different pretreatment methods
预处理方法 | F-measure/% | Accuracy/% |
---|
Raw | 53.93 | 56.84 | Smooth | 53.93 | 56.84 | SNV | 53.93 | 56.84 | MSC | 54.02 | 57.89 | SG 1st-Der | 80.90 | 82.11 | WT | 68.57 | 71.05 |
|
Table 4. NBC model results of qualitative identification of ‘Yali' pears with different pretreatment methods
预处理方法 | F-measure/% | Accuracy/% |
---|
Raw | 85.19 | 87.37 | Smooth | 84.81 | 87.37 | SNV | 85.19 | 87.37 | MSC | 85.19 | 87.37 | SG 1st-Der | 56.52 | 68.42 | WT | 90.24 | 91.58 |
|
Table 5. SVM model results of qualitative identification of ‘Yali' pears with different pretreatment methods
预处理方法 | F-measure/% | Accuracy/% |
---|
Raw | 84.21 | 87.37 | Smooth | 84.42 | 87.37 | SNV | 85.88 | 87.37 | MSC | 83.23 | 85.79 | SG 1st-Der | 77.19 | 79.47 | WT | 91.46 | 92.63 |
|
Table 6. AdaBoost qualitative identification results of ‘Yali' pears with different pretreatment methods
分类模型 | F-measure /% | Accuracy /% | 预测时间 估算*/s |
---|
WT-kNN | 79.52 | 82.11 | 0.04 | SG 1st-Der-NBC | 75.68 | 81.05 | 0.03 | WT-SVM | 88.61 | 90.53 | 0.04 | WT-AdaBoost | 90.91 | 92.63 | 0.12 |
|
Table 7. KNN, NBC, SVM and AdaBoost model test set prediction results