Author Affiliations
Hunan Key Laboratory of Electric Power Robot, College of Electrical and Information Engineering, Changsha University of Science & Technology, Changsha, Hunan 410114, Chinashow less
Fig. 1. Flow chart of instrumental detection and reading recognition
Fig. 2. Dehazing results under different D. (a) Hazy image; (b) image before mapping; (c) dehazing result under D=1.9; (d) dehazing result under D=2.3; (e) dehazing result under D=2.7
Fig. 3. Structural diagrams of ResNet network. (a) Residual unit; (b) improved residual unit; (c) our deraining network
Fig. 4. Comparison of image pixel distributions. (a) Histogram of image; (b) histogram of detail layer
Fig. 5. Extraction process of pointer. (a) Dial positioning; (b) dial correction and pointer extraction
Fig. 6. Comparison of dehazing images. (a) Hazy images; (b) dehazing images using dark channel; (c) dehazing images using MSRCR; (d) our dehazing images
Fig. 7. Contrast ofderaining images. (a)Rainy images; (b)deraining images without filter layer; (c) our deraining images
Fig. 8. Contrast of separation results of pointer area. (a1)(a2) Pointer area; (b1)(b2) adaptive Otsu algorithm; (c1)(c2) iterative threshold algorithm; (d1)(d2) our algorithm
Image No. | Hazy image | Dark channel dehazing method | MSRCR dehazing method | Our dehazing method |
---|
Entropy | MSE | Entropy | MSE | Entropy | MSE | Entropy | MSE |
---|
Image 1 | 6.74 | 6356 | 12.09 | 1221 | 11.94 | 2211 | 14.04 | 1167 | Image 2 | 8.47 | 5019 | 12.42 | 2111 | 11.86 | 2185 | 14.12 | 1812 | Image 3 | 10.21 | 2828 | 12.38 | 3423 | 11.69 | 2326 | 13.84 | 2136 | Image 4 | 11.24 | 2256 | 12.43 | 3909 | 11.56 | 2880 | 13.76 | 2474 |
|
Table 1. Comparison of image entropy and MSE
Method | 225 pixel×225 pixel | 415 pixel×415 pixel |
---|
Dark channel dehazing method | 1.50 | 1.55 | MSRCR dehazing method | 2.20 | 2.11 | Our method | 1.02 | 1.02 |
|
Table 2. Dehazing process time comparison of images unit: s
Evaluating indicator (average) | Rainy image | Deraining image without filter layer | Our deraining image |
---|
PSNR | 20.13 | 30.16 | 35.02 | SSIM | 0.58 | 0.89 | 0.93 |
|
Table 3. Comparison of image PSNR and SSIM
Manual recognition method | Adaptive Otsu algorithm | Our method | Error |
---|
0.5° | 0.56° | 0.52° | +0.02° | 6.0° | 6.30° | 6.23° | +0.23° | 9.5° | 9.25° | 9.35° | -0.15° | 26.5° | 26.75° | 26.0° | -0.50° | 32.0° | 31.4° | 31.7° | -0.30° |
|
Table 4. Angle recognition results of meter reading