Author Affiliations
School of Information Engineering, Zhejiang A & F University, Lin'an, Zhejiang 311300, Chinashow less
Fig. 1. Typical defective samples
Fig. 2. Flow chart of recognition algorithm
Fig. 3. Illustration of image block
Fig. 4. Original images. (a) Original image of defect of rotary veneer; (b) original image of defect of rotary veneer with texture
Fig. 5. Grayscale images. (a) Grayscale image of rotary veneer; (b) grayscale of rotary veneer with texture
Fig. 6. Grayscale histograms of original images. (a) Grayscale histogram of rotary veneer; (b) grayscale histogram of rotary veneer with texture
Fig. 7. Comparison of gray histograms of normal block and knot block. (a) Gray histogram of normal block of rotary veneer; (b) gray histogram of knot block of rotary veneer; (c) gray histogram of normal block of rotary veneer with texture; (d) gray histogram of knot block of rotary veneer with texture
Fig. 8. Preliminary identification results. (a) Preliminary identification result of rotary veneer; (b) preliminary identification result of rotary veneer with texture
Fig. 9. Parameter optimization result
Fig. 10. Final recognition results. (a) Rotary veneer; (b) rotary veneer with texture
Item | True flaw | True normal |
---|
Predicted defect | TP | FP | Predicted normal | FN | TN |
|
Table 1. Four combinations of true and false predictions
Method | TPR | NPV | ACC |
---|
Grayscale histogram | 100 | 82.6 | 96.1 | Textural feature | 100 | 73.9 | 94.0 | Proposed method | 100 | 89.5 | 97.0 |
|
Table 2. Statistics data of rotary veneers%
Method | TPR | NPV | ACC |
---|
Grayscale histogram | 100 | 90.9 | 93.80 | Textural feature | 92 | 78.8 | 88.90 | Proposed method | 92 | 95.6 | 97.20 |
|
Table 3. Statistics data of rotary veneers with texture%