Hyperspectral Image Classification Based on Hyperpixel Segmentation and Convolutional Neural Network

Rujun Chen; Yunwei Pu; Fengzhen Wu; Yuceng Liu; Qi Li

doi:10.3788/LOP222551

Journals >Laser & Optoelectronics Progress >Volume 60 >Issue 16 >Page 1610010 > Article

Laser & Optoelectronics Progress
Vol. 60, Issue 16, 1610010 (2023)

Hyperspectral Image Classification Based on Hyperpixel Segmentation and Convolutional Neural Network

Rujun Chen¹, Yunwei Pu^1,2,*, Fengzhen Wu¹, Yuceng Liu¹, and Qi Li¹

Author Affiliations

¹Faculty of Land Resources Engineering, Kunming University of Science and Technology, Kunming 650093, Yunnan, China

²Computing Center, Kunming University of Science and Technology, Kunming 650500, Yunnan, China

show less

DOI: 10.3788/LOP222551 Cite this Article Set citation alerts

Rujun Chen, Yunwei Pu, Fengzhen Wu, Yuceng Liu, Qi Li. Hyperspectral Image Classification Based on Hyperpixel Segmentation and Convolutional Neural Network[J]. Laser & Optoelectronics Progress, 2023, 60(16): 1610010 Copy Citation Text

show less

Fig. 1. Proposed CNN

Download full size

Fig. 2. Pseudo-color image and ground object truth map of the WHU-Hi-Longkou dataset. (a) Pseudo-color image; (b) real image

Download full size

Fig. 3. Pseudo-color image and ground object truth map of the WHU-Hi-HongHu dataset. (a) Pseudo-color image; (b) real image

Download full size

Fig. 4. Spectral information segmentation results on WHU-Hi-Longkou dataset. (a) KNN; (b) SVM; (c) RF; (d) CNN; (e) ground truth

Download full size

Fig. 5. Spectral information segmentation results on WHU-Hi-HongHu dataset. (a) KNN; (b) SVM; (c) RF; (d) CNN; (e) ground truth

Download full size

Fig. 6. Segmentation results of different hyperpixel algorithms on WHU-Hi-Longkou dataset. (a) SLICO; (b) PCA-SLICO; (c) SLIC; (d) PCA-SLIC; (e) MSLIC; (f) PCA-MSLIC

Download full size

Fig. 7. Segmentation results of different hyperpixel algorithms on WHU-Hi-HongHu dataset. (a) SLICO; (b) PCA-SLICO; (c) SLIC; (d) PCA-SLIC; (e) MSLIC; (f) PCA-MSLIC

Download full size

Fig. 8. Classification results of different methods on WHU-Hi-Longkou dataset. (a) PMS-KNN; (b) PMS-SVM; (c) PMS-RF;(d) PMS-CNN; (e) grouth truth

Download full size

Fig. 9. Classification results of different methods on WHU-Hi-Longkou dataset. (a) PMS-KNN; (b) PMS-SVM; (c) PMS-RF;(d) PMS-CNN; (e) grouth truth

Download full size

Category	KNN	SVM	RF	CNN
Corn	94.93	98.41	88.53	98.45
Cotton	76.80	92.02	76.80	95.45
Sesame	82.96	98.25	82.21	98.70
Round leaf soybean	69.00	89.60	57.16	90.83
Long leaf soybean	88.30	96.05	69.80	97.91
Rice	99.24	99.48	64.20	98.68
Wave	99.87	99.88	92.18	99.83
Houses and roads	87.82	95.37	82.34	94.58
Mixed weeds	85.28	96.13	74.18	97.41
OA /%	89.88	96.77	82.51	97.18
Kappa coefficient /%	86.97	95.78	78.00	96.31

Table 1. CA of different algorithms on WHU-Hi-Longkou dataset

Category	KNN	SVM	RF	CNN
Roof	84.37	76.02	78.67	92.17
Road	81.97	83.56	83.39	78.81
Exposed soil	63.15	74.45	65.50	73.66
Cotton	49.90	67.64	62.75	87.36
Cotton wood	60.21	78.70	68.05	64.24
Rape	66.54	81.40	72.56	92.47
Chinese cabbage	45.33	53.10	46.34	49.05
Pakchoi	38.80	39.56	20.93	58.84
Cabbage	88.38	89.18	84.81	88.28
Mustard tuber	38.70	47.20	40.47	66.60
Cauliflower	31.90	40.22	30.36	60.92
Green vegetables	56.95	54.81	54.33	61.74
Small green vegetables	45.56	49.12	52.04	71.80
Asparagus lettuce	61.54	56.82	44.33	50.75
Lettuce	87.30	80.42	73.91	82.85
Film is covered with lettuce	76.67	69.85	80.14	82.85
Thin film covered with lettuce	78.53	77.82	70.30	75.83
Carrot	68.37	73.40	68.31	84.24
Ternip	68.68	75.56	71.66	79.01
Garlic bolt	77.13	80.36	76.20	76.94
Bean	71.18	71.10	76.68	89.47
Persimmon tree	64.02	77.34	74.47	87.46
OA /%	63.02	72.68	68.26	85.43
Kappa coefficient /%	56.78	67.04	62.02	82.10

Table 2. CA of different algorithms on WHU-Hi-HongHu dataset

Category	PMS-KNN	PMS-SVM	PMS-RF	PMS-CNN
Corn	97.74	99.64	94.37	99.85
Cotton	83.53	96.03	66.15	99.94
Sesame	89.90	98.06	83.03	98.35
Round leaf soybean	76.57	92.52	70.88	99.08
Long leaf soybean	93.84	97.02	76.10	99.15
Rice	99.62	9.540	99.08	99.67
Wave	99.88	99.87	99.85	99.72
Houses and roads	91.15	94.75	88.04	95.36
Mixed weeds	86.41	97.64	81.50	97.61
OA /%	92.76	97.80	89.35	99.45
Kappa coefficient /%	90.63	97.10	86.30	99.27

Table 3. CA on WHU-Hi-Longkou dataset

Category	PMS-KNN	PMS-SVM	PMS-RF	PMS-CNN
Roof	95.62	80.33	93.58	98.38
Road	94.60	83.67	87.62	94.02
Exposed soil	87.36	91.36	86.13	96.33
Cotton	87.54	84.46	82.00	98.10
Cotton wood	92.34	84.39	91.18	99.45
Rape	87.99	87.79	86.40	96.57
Chinese cabbage	74.94	65.26	72.08	92.17
Pakchoi	86.60	17.60	77.20	99.95
Cabbage	96.28	91.77	94.73	97.60
Mustard tuber	73.43	66.86	76.93	96.85
Cauliflower	78.04	59.00	87.98	95.36
Green vegetables	86.78	56.75	85.84	96.27
Small green vegetables	74.96	73.16	76.14	97.06
Asparagus lettuce	86.90	64.80	87.98	97.22
Lettuce	94.66	67.80	99.00	100
Film is covered with lettuce	94.95	85.58	86.55	99.10
Thin film covered with lettuce	95.20	60.40	96.84	100
Carrot	94.42	78.54	95.64	99.40
Ternip	87.04	81.54	86.51	95.24
Garlic bolt	97.44	80.70	96.73	98.58
Bean	99.70	0	99.70	99.70
Persimmon tree	95.85	76.86	99.23	99.24
OA /%	88.97	82.90	86.84	97.60
Kappa coefficient /%	86.30	78.67	83.80	97.07