Remote Sensing Scene Image Classification Based on mmsCNN–HMM with Stacking Ensemble Model
The development of convolution neural networks (CNNs) has become a significant means to solve the problem of remote sensing scene image classification. However, well-performing CNNs generally have high complexity and are prone to overfitting. To handle the above problem, we present a new classificat...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-09-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/14/17/4423 |
| _version_ | 1827665035726422016 |
|---|---|
| author | Xiang Cheng Hong Lei |
| author_facet | Xiang Cheng Hong Lei |
| author_sort | Xiang Cheng |
| collection | DOAJ |
| description | The development of convolution neural networks (CNNs) has become a significant means to solve the problem of remote sensing scene image classification. However, well-performing CNNs generally have high complexity and are prone to overfitting. To handle the above problem, we present a new classification approach using an mmsCNN–HMM combined model with stacking ensemble mechanism in this paper. First of all, a modified multi-scale convolution neural network (mmsCNN) is proposed to extract multi-scale structural features, which has a lightweight structure and can avoid high computational complexity. Then, we utilize a hidden Markov model (HMM) to mine the context information of the extracted features of the whole sample image. For different categories of scene images, the corresponding HMM is trained and all the trained HMMs form an HMM group. In addition, our approach is based on a stacking ensemble learning scheme, in which the preliminary predicted values generated by the HMM group are used in an extreme gradient boosting (XGBoost) model to generate the final prediction. This stacking ensemble learning mechanism integrates multiple models to make decisions together, which can effectively prevent overfitting while ensuring accuracy. Finally, the trained XGBoost model conducts the scene category prediction. In this paper, the six most widely used remote sensing scene datasets, UCM, RSSCN, SIRI-WHU, WHU-RS, AID, and NWPU, are selected to carry out all kinds of experiments. The numerical experiments verify that the proposed approach shows more important advantages than the advanced approaches. |
| first_indexed | 2024-03-10T01:16:58Z |
| format | Article |
| id | doaj.art-423ce13fb0d94f0c9ed12a5766f279e3 |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-03-10T01:16:58Z |
| publishDate | 2022-09-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-423ce13fb0d94f0c9ed12a5766f279e32023-11-23T14:06:19ZengMDPI AGRemote Sensing2072-42922022-09-011417442310.3390/rs14174423Remote Sensing Scene Image Classification Based on mmsCNN–HMM with Stacking Ensemble ModelXiang Cheng0Hong Lei1Department of Space Microwave Remote Sensing System, Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100190, ChinaDepartment of Space Microwave Remote Sensing System, Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100190, ChinaThe development of convolution neural networks (CNNs) has become a significant means to solve the problem of remote sensing scene image classification. However, well-performing CNNs generally have high complexity and are prone to overfitting. To handle the above problem, we present a new classification approach using an mmsCNN–HMM combined model with stacking ensemble mechanism in this paper. First of all, a modified multi-scale convolution neural network (mmsCNN) is proposed to extract multi-scale structural features, which has a lightweight structure and can avoid high computational complexity. Then, we utilize a hidden Markov model (HMM) to mine the context information of the extracted features of the whole sample image. For different categories of scene images, the corresponding HMM is trained and all the trained HMMs form an HMM group. In addition, our approach is based on a stacking ensemble learning scheme, in which the preliminary predicted values generated by the HMM group are used in an extreme gradient boosting (XGBoost) model to generate the final prediction. This stacking ensemble learning mechanism integrates multiple models to make decisions together, which can effectively prevent overfitting while ensuring accuracy. Finally, the trained XGBoost model conducts the scene category prediction. In this paper, the six most widely used remote sensing scene datasets, UCM, RSSCN, SIRI-WHU, WHU-RS, AID, and NWPU, are selected to carry out all kinds of experiments. The numerical experiments verify that the proposed approach shows more important advantages than the advanced approaches.https://www.mdpi.com/2072-4292/14/17/4423remote sensing scene image classificationdeep learningCNNhidden Markov model (HMM) |
| spellingShingle | Xiang Cheng Hong Lei Remote Sensing Scene Image Classification Based on mmsCNN–HMM with Stacking Ensemble Model Remote Sensing remote sensing scene image classification deep learning CNN hidden Markov model (HMM) |
| title | Remote Sensing Scene Image Classification Based on mmsCNN–HMM with Stacking Ensemble Model |
| title_full | Remote Sensing Scene Image Classification Based on mmsCNN–HMM with Stacking Ensemble Model |
| title_fullStr | Remote Sensing Scene Image Classification Based on mmsCNN–HMM with Stacking Ensemble Model |
| title_full_unstemmed | Remote Sensing Scene Image Classification Based on mmsCNN–HMM with Stacking Ensemble Model |
| title_short | Remote Sensing Scene Image Classification Based on mmsCNN–HMM with Stacking Ensemble Model |
| title_sort | remote sensing scene image classification based on mmscnn hmm with stacking ensemble model |
| topic | remote sensing scene image classification deep learning CNN hidden Markov model (HMM) |
| url | https://www.mdpi.com/2072-4292/14/17/4423 |
| work_keys_str_mv | AT xiangcheng remotesensingsceneimageclassificationbasedonmmscnnhmmwithstackingensemblemodel AT honglei remotesensingsceneimageclassificationbasedonmmscnnhmmwithstackingensemblemodel |