Laser & Optoelectronics Progress, Volume. 57, Issue 20, 201101(2020)
GGCN: GPU-Based Hyperspectral Image Classification Algorithm
Figures & Tables(12)
Table 1. Algorithm pseudocode
View tableTable 1. Algorithm pseudocode
Algorithm:GGCN Input: Hyperspectral image 1, Data preprocessing: processing << i-th iteration: Forward propagation 2, Convolutional: convol <<< gridsize, blocksize, 0, stream>>> 3, Pooling: maxpooling <<< gridsize, blocksize, 0, stream>>> 4, Fully connected: fullyconnected <<< gridsize, blocksize, 0, stream>>> 5, Output: output << 6, Copy classification results to CPU to calculate the loss: 7, Copy data: cudaMemcpy() 8, Calculate the loss: lossfunction() Backward propagation 9, Output: bp_output <<< gridsize, blocksize, 0, stream>>> 10, Fully Connected: bp_fullyconnected<<< gridsize, blocksize,0,stream>>> 11, Pooling: bp_maxpooling << 12, Convolutional: bp_update_kernel <<< gridsize, blocksize, 0, stream>>> Output End
Table 2. Information of the remote sensing datasets
View tableTable 2. Information of the remote sensing datasets
Dataset Sensor Class number Dimension Top 5 classes Size /MB KSC AVIRIS 13 512 × 614×176 Water, scrub, spartna-marsh, mud-flats, salt-marsh 56.8 PU POSIS 9 610×340×103 Meadows, asphalt, bare-soil, self-blocking bricks, trees 33.2 Indian Pines AVIRIS 16 145×145×224 Soybean-mintill, corn-notill, woods, soybean-notill, corn-mintill 5.7
Table 3. Comparison of time consumption of different data preprocessing methods
View tableTable 3. Comparison of time consumption of different data preprocessing methods
Dataset Neighbor pixel extract strategy Time /s CPU PNPE G-PNPE KSC 1P 4N 8N 2.65 4.89 6.21 0.45 0.88 1.12 0.51 0.92 1.22 PU 1P 4N 8N 2.21 3.30 3.87 0.31 0.52 0.66 0.33 0.49 0.71 Indian Pines 1P 4N 8N 1.05 1.65 2.17 0.17 0.21 0.28 0.18 0.20 0.26
Table 4. Comparison of running time and speedup of different classification models
View tableTable 4. Comparison of running time and speedup of different classification models
Dataset Method Time /s Speedup ratio MBGD(batchsize is 10) MBGD(batchsize is 100) KSC Cube-CNN-SVM GCN GGCN 23123.62 3487.34 2834.01 23012.49 3322.34 2598.78 1.0 6.6 8.2 PU Cube-CNN-SVM GCN GGCN 2231.23 351.46 286.22 2187.75 338.11 230.06 1.0 6.3 7.8 Indian Pines Cube-CNN-SVM GCN GGCN 453.62 107.34 84.01 422.49 102.34 80.78 1.0 4.2 5.1
Table 5. Accuracy of different classification models
View tableTable 5. Accuracy of different classification models
Dataset Method Accuracy /% MBGD(batchsize is 10) MBGD(batchsize is 100) KSC Cube-CNN-SVM GCN GGCN 93.78 93.33 93.67 93.47 93.12 93.92 PU Cube-CNN-VM GCN GGCN 96.67 96.23 96.34 95.21 95.61 95.69 Indian Pines Cube-CNN-SVM GCN GGCN 94.78 94.73 94.87 94.67 94.52 94.42
Table 6. Ratio of time between the improved model and the original model
View tableTable 6. Ratio of time between the improved model and the original model
Layer Percentage /% GCN GGCN Preprocessing 1.0 2.4 Convolution 38.2 28.0 Pooling 2.4 5.6 Fully connection 27.1 24.0 Output 19.0 22.0 Others 13.3 18.0
Tools
Get Citation
Copy Citation Text
Minghua Zhang, Yaqing Zou, Wei Song, Dongmei Huang, Zhixiang Liu. GGCN: GPU-Based Hyperspectral Image Classification Algorithm[J]. Laser & Optoelectronics Progress, 2020, 57(20): 201101
Paper Information
Category: Imaging Systems
Received: Dec. 16, 2019
Accepted: Feb. 25, 2020
Published Online: Oct. 13, 2020
The Author Email: Huang Dongmei (dmhuang@shou.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20