Yili Zhao
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Yili Zhao has a total of 24 co-authors. Here are the top co-authors by number of shared publications:
- Leiguang Wang
College of landscape Architecture and Horticulture, Southwest Forestry University, Kunming, China
4 shared publications - Qinling Dai
Art and Design College, Southwest Forestry University, Kunming 650024, China
2 shared publications - Dan Xu
Yunnan University
1 shared publication - Haoyu Fu
The College of Soil and Water Conservation, Southwest Forestry University, Kunming 650224, China
1 shared publication Hua Zhou
Southwest Forestry University
1 shared publication- Hua Zhou
College of Big Data and Intelligent Engineering, Southwest Forestry University, Kunming 650224, China
1 shared publication
Latest Publications
- Open Access
- 28 May 2025
- Published by MDPI AG in Remote Sensing
- Vol. 17 (11) , 1880
Remote Sensing Image Segmentation Network That Integrates Global–Local Multi-Scale Information with Deep and Shallow Features
As the spatial resolution of remote sensing images continues to increase, the complexity of the information they carry also grows. Remote sensing images are characterized by large imaging areas, scattered distributions of similar objects, intricate boundary shapes, and a high density of small objects, all of which pose significant challenges for semantic segmentation tasks. To address these challenges, we propose a Remote Sensing Image Segmentation Network that Integrates Global–Local Multi-Scale Information with Deep and Shallow Features (GLDSFNet). To better handle the wide variations in object sizes and complex boundary shapes, we design a Global–Local Multi-Scale Feature Fusion Module (GLMFM) that enhances segmentation performance by fully leveraging multi-scale information and global context. Additionally, to improve the segmentation of small objects, we propose a Shallow–Deep Feature Fusion Module (SDFFM), which effectively integrates deep semantic information with shallow spatial features through mutual guidance, retaining the advantages of both. Extensive ablation and comparative experiments conducted on two public remote sensing datasets, ISPRS Vaihingen and Potsdam, demonstrate that our proposed GLDSFNet outperforms state-of-the-art methods.
- 28 May 2025
- Published by MDPI AG in Remote Sensing
- Vol. 17 (11) , 1880
Open AccessRemote Sensing Image Segmentation Network That Integrates Global–Local Multi-Scale Information with Deep and Shallow Features
As the spatial resolution of remote sensing images continues to increase, the complexity of the information they carry also grows. Remote sensing images are characterized by large imaging areas, scattered distributions of similar objects, intricate boundary shapes, and a high density of small objects, all of which pose significant challenges for semantic segmentation tasks. To address these challenges, we propose a Remote Sensing Image Segmentation Network that Integrates Global–Local Multi-Scale Information with Deep and Shallow Features (GLDSFNet). To better handle the wide variations in object sizes and complex boundary shapes, we design a Global–Local Multi-Scale Feature Fusion Module (GLMFM) that enhances segmentation performance by fully leveraging multi-scale information and global context. Additionally, to improve the segmentation of small objects, we propose a Shallow–Deep Feature Fusion Module (SDFFM), which effectively integrates deep semantic information with shallow spatial features through mutual guidance, retaining the advantages of both. Extensive ablation and comparative experiments conducted on two public remote sensing datasets, ISPRS Vaihingen and Potsdam, demonstrate that our proposed GLDSFNet outperforms state-of-the-art methods.
- 1 January 2025
- Published by SPIE-Intl Soc Optical Eng in Journal of Electronic Imaging
- Vol. 34 (01) , 013024-11
TransUNetFormer: let hybrid convolutional neural network + transformer encoder and decoder provide powerful support for remote sensing image segmentation
We focus on the complexity and multi-scale importance of semantic segmentation of remote sensing images, a fundamental task in earth science research. We propose an architecture, TransUNetFormer, that is potent such as U-Net designed that extensively integrates convolutional neural network (CNN) + transformer fusion in both the encoder and decoder; this fusion emphasizes the significance of global contextual information and local feature details. TransUNetFormer achieves superior generalization for remote sensing image segmentation, particularly in capturing multi-scale features within its encoder-decoder architecture. The encoder incorporates design principles inspired by TransUNet, leveraging a CNN + transformer component for an efficient hybrid. In addition, a CNN + transformer hybrid block in the decoder, DP-hybrid, efficiently captures rich global-local features at each upsampling step. We introduce a fusion-concatenation module to dynamically generate weights during the interaction between the encoder and decoder, facilitating feature map fusion. Finally, an efficient feature refinement segmentation head is devised to purify shallow-stage encoder features and optimize the most profound global-local features in the decoder for fusion output. Experimental results on two widely used datasets, ISPRS Potsdam and LoveDA Urban, show the effectiveness and potential of TransUNetFormer. To our knowledge, this is the first hybrid CNN + transformer network specifically designed for remote sensing image segmentation.
- 1 January 2025
- Published by SPIE-Intl Soc Optical Eng in Journal of Electronic Imaging
- Vol. 34 (01) , 013024-11
TransUNetFormer: let hybrid convolutional neural network + transformer encoder and decoder provide powerful support for remote sensing image segmentation
We focus on the complexity and multi-scale importance of semantic segmentation of remote sensing images, a fundamental task in earth science research. We propose an architecture, TransUNetFormer, that is potent such as U-Net designed that extensively integrates convolutional neural network (CNN) + transformer fusion in both the encoder and decoder; this fusion emphasizes the significance of global contextual information and local feature details. TransUNetFormer achieves superior generalization for remote sensing image segmentation, particularly in capturing multi-scale features within its encoder-decoder architecture. The encoder incorporates design principles inspired by TransUNet, leveraging a CNN + transformer component for an efficient hybrid. In addition, a CNN + transformer hybrid block in the decoder, DP-hybrid, efficiently captures rich global-local features at each upsampling step. We introduce a fusion-concatenation module to dynamically generate weights during the interaction between the encoder and decoder, facilitating feature map fusion. Finally, an efficient feature refinement segmentation head is devised to purify shallow-stage encoder features and optimize the most profound global-local features in the decoder for fusion output. Experimental results on two widely used datasets, ISPRS Potsdam and LoveDA Urban, show the effectiveness and potential of TransUNetFormer. To our knowledge, this is the first hybrid CNN + transformer network specifically designed for remote sensing image segmentation.
- 17 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 9108-9113
Build General Pansharpening Model Through a Simple Sensor-Specific Model Selection Strategy
Deep-learning-based pansharpening methods have become a hot field in the remote sensing community. However, most existing deep learning pansharpening methods are conducted in a sensor-specific and supervised way. That is, pairs of panchromatic (Pan) and multispectral (MS) images acquired from a given sensor are fed into networks. The trained model can achieve a satisfactory fusion result only if the images are fused from the same sensor as the training images. The trained model from one sensor cannot be transferred to another. How to train a general model suitable for different sensors is a challenging problem. In this context, we propose a simple sensor-type detecting strategy, which can be integrated with any sensor-specific pansharpening methods and result in a general fusion framework across different sensors. More specifically, for a given deep-learning-based pansharpening method, sufficient Pan and MS image pairs from multiple sensors are fed into the network respectively, and models from multiple sensors are obtained to construct a model pool. Thereafter, a pair of local patches of MS and Pan from the test images are fed into existing sensor-specific models in the pool respectively. The pansharpening quality of outputs are evaluated with multiple fusing indices, such as quality without reference index (QNR). The model that achieves the best performance in terms of index is selected as the candidate model and further employs the images over the whole scene. To evaluate the effectiveness of the proposed framework, WSDFNet, a state-of-the-art pansharpening method was trained with GaoFen-1, QuickBird, and WorldView-3 datasets respectively. The QNR index is the optimal index for predicting the sensor type. With 100 images from three datasets as testing images, the correct accuracy rate of the statistical QNR index on sensor differentiation performance is 86, 69, and 62, respectively, which indicates this model is promising for distinguishing the sensor type. This proposed prediction strategy is simple and flexible, and it can be combined with existing deep-learning methods to form a complete fusion process without distinguishing the sensor type.
- 17 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 9108-9113
Build General Pansharpening Model Through a Simple Sensor-Specific Model Selection Strategy
Deep-learning-based pansharpening methods have become a hot field in the remote sensing community. However, most existing deep learning pansharpening methods are conducted in a sensor-specific and supervised way. That is, pairs of panchromatic (Pan) and multispectral (MS) images acquired from a given sensor are fed into networks. The trained model can achieve a satisfactory fusion result only if the images are fused from the same sensor as the training images. The trained model from one sensor cannot be transferred to another. How to train a general model suitable for different sensors is a challenging problem. In this context, we propose a simple sensor-type detecting strategy, which can be integrated with any sensor-specific pansharpening methods and result in a general fusion framework across different sensors. More specifically, for a given deep-learning-based pansharpening method, sufficient Pan and MS image pairs from multiple sensors are fed into the network respectively, and models from multiple sensors are obtained to construct a model pool. Thereafter, a pair of local patches of MS and Pan from the test images are fed into existing sensor-specific models in the pool respectively. The pansharpening quality of outputs are evaluated with multiple fusing indices, such as quality without reference index (QNR). The model that achieves the best performance in terms of index is selected as the candidate model and further employs the images over the whole scene. To evaluate the effectiveness of the proposed framework, WSDFNet, a state-of-the-art pansharpening method was trained with GaoFen-1, QuickBird, and WorldView-3 datasets respectively. The QNR index is the optimal index for predicting the sensor type. With 100 images from three datasets as testing images, the correct accuracy rate of the statistical QNR index on sensor differentiation performance is 86, 69, and 62, respectively, which indicates this model is promising for distinguishing the sensor type. This proposed prediction strategy is simple and flexible, and it can be combined with existing deep-learning methods to form a complete fusion process without distinguishing the sensor type.
- Open Access
- 27 October 2023
- Published by MDPI AG in Remote Sensing
- Vol. 15 (21) , 5148
Boosting Semantic Segmentation of Remote Sensing Images by Introducing Edge Extraction Network and Spectral Indices
Deep convolutional neural networks have greatly enhanced the semantic segmentation of remote sensing images. However, most networks are primarily designed to process imagery with red, green, and blue bands. Although it is feasible to directly utilize established networks and pre-trained models for remotely sensed images, they suffer from imprecise land object contour localization and unsatisfactory segmentation results. These networks still need to explore the domain knowledge embedded in images. Therefore, we boost the segmentation performance of remote sensing images by augmenting the network input with multiple nonlinear spectral indices, such as vegetation and water indices, and introducing a novel holistic attention edge detection network (HAE-RNet). Experiments were conducted on the GID and Vaihingen datasets. The results showed that the NIR-NDWI/DSM-GNDVI-R-G-B (6C-2) band combination produced the best segmentation results for both datasets. The edge extraction block benefits better contour localization. The proposed network achieved a state-of-the-art performance in both the quantitative evaluation and visual inspection.
- 27 October 2023
- Published by MDPI AG in Remote Sensing
- Vol. 15 (21) , 5148
Open AccessBoosting Semantic Segmentation of Remote Sensing Images by Introducing Edge Extraction Network and Spectral Indices
Deep convolutional neural networks have greatly enhanced the semantic segmentation of remote sensing images. However, most networks are primarily designed to process imagery with red, green, and blue bands. Although it is feasible to directly utilize established networks and pre-trained models for remotely sensed images, they suffer from imprecise land object contour localization and unsatisfactory segmentation results. These networks still need to explore the domain knowledge embedded in images. Therefore, we boost the segmentation performance of remote sensing images by augmenting the network input with multiple nonlinear spectral indices, such as vegetation and water indices, and introducing a novel holistic attention edge detection network (HAE-RNet). Experiments were conducted on the GID and Vaihingen datasets. The results showed that the NIR-NDWI/DSM-GNDVI-R-G-B (6C-2) band combination produced the best segmentation results for both datasets. The edge extraction block benefits better contour localization. The proposed network achieved a state-of-the-art performance in both the quantitative evaluation and visual inspection.
- Open Access
- 17 October 2022
- Published by Springer Nature in Applied Intelligence
- Vol. 53 (11) , 13852-13866
Improved generalization performance of convolutional neural networks with LossDA
In recent years, convolutional neural networks (CNNs) have been used in many fields. Nowadays, CNNs have a high learning capability, and this learning capability is accompanied by a more complex model architecture. Complex model architectures allow CNNs to learn more data features, but such a learning process tends to reduce the training model’s ability to generalize to unknown data, and may be associated with problems of overfitting. Although many regularization methods have been proposed, such as data augmentation, batch normalization, and Dropout, research on improving generalization performance is still a common concern in the training process of robust CNNs. In this paper, we propose a dynamically controllable adjustment method, which we call LossDA, that embeds a disturbance variable in the fully-connected layer. The trend of this variable is kept consistent with the training loss, while the magnitude of the variable can be preset to adapt to the training process of different models. Through this dynamic adjustment, the training process of CNNs can be adaptively adjusted. The whole regularization process can improve the generalization performance of CNNs while helping to suppress overfitting. To evaluate this method, this paper conducts comparative experiments on MNIST, FashionMNIST, CIFAR-10, Cats_vs_Dogs, and miniImagenet datasets. The experimental results show that the method can improve the model performance of Light CNNs and Transfer CNNs (InceptionResNet, VGG19, ResNet50, and InceptionV3). The average maximum improvement in accuracy of Light CNNs is 4.62%, F1 is 3.99%, and Recall is 4.69%. The average maximum improvement accuracy of Transfer CNNs is 4.17%, F1 is 5.64%, and Recall is 4.05%.
- 17 October 2022
- Published by Springer Nature in Applied Intelligence
- Vol. 53 (11) , 13852-13866
Open AccessImproved generalization performance of convolutional neural networks with LossDA
In recent years, convolutional neural networks (CNNs) have been used in many fields. Nowadays, CNNs have a high learning capability, and this learning capability is accompanied by a more complex model architecture. Complex model architectures allow CNNs to learn more data features, but such a learning process tends to reduce the training model’s ability to generalize to unknown data, and may be associated with problems of overfitting. Although many regularization methods have been proposed, such as data augmentation, batch normalization, and Dropout, research on improving generalization performance is still a common concern in the training process of robust CNNs. In this paper, we propose a dynamically controllable adjustment method, which we call LossDA, that embeds a disturbance variable in the fully-connected layer. The trend of this variable is kept consistent with the training loss, while the magnitude of the variable can be preset to adapt to the training process of different models. Through this dynamic adjustment, the training process of CNNs can be adaptively adjusted. The whole regularization process can improve the generalization performance of CNNs while helping to suppress overfitting. To evaluate this method, this paper conducts comparative experiments on MNIST, FashionMNIST, CIFAR-10, Cats_vs_Dogs, and miniImagenet datasets. The experimental results show that the method can improve the model performance of Light CNNs and Transfer CNNs (InceptionResNet, VGG19, ResNet50, and InceptionV3). The average maximum improvement in accuracy of Light CNNs is 4.62%, F1 is 3.99%, and Recall is 4.69%. The average maximum improvement accuracy of Transfer CNNs is 4.17%, F1 is 5.64%, and Recall is 4.05%.
- Open Access
A Research on the Sharing Platform of Wild Bird Data in Yunnan Province Based on Blockchain and Interstellar File System
Sharing scientific data is an effective means to rationally exploit scientific data and is vital to promote the development of the industrial chain and improve the level of science and technology. In recent years, the popularity of the open data platform has increased, but problems remain, including imperfect system architecture, unsound privacy and security, and non-standardized interaction data. To address these problems, the blockchain’s decentralization, smart contracts, distributed storage, and other features can be used as the core technology for open data systems. This paper addresses the problems of opening, allocation-right confirmation, sharing, and rational use of wild-bird data from Yunnan Province, China. A data storage model is proposed based on the blockchain and interstellar file system and is applied to wild-bird data to overcome the mutual distrust between ornithology institutions in the collaborative processing and data storage of bird data. The model provides secure storage and secure access control of bird data in the cloud, thereby ensuring the decentralized and secure storage of wild-bird data for multiple research institutions.
Open AccessA Research on the Sharing Platform of Wild Bird Data in Yunnan Province Based on Blockchain and Interstellar File System
Sharing scientific data is an effective means to rationally exploit scientific data and is vital to promote the development of the industrial chain and improve the level of science and technology. In recent years, the popularity of the open data platform has increased, but problems remain, including imperfect system architecture, unsound privacy and security, and non-standardized interaction data. To address these problems, the blockchain’s decentralization, smart contracts, distributed storage, and other features can be used as the core technology for open data systems. This paper addresses the problems of opening, allocation-right confirmation, sharing, and rational use of wild-bird data from Yunnan Province, China. A data storage model is proposed based on the blockchain and interstellar file system and is applied to wild-bird data to overcome the mutual distrust between ornithology institutions in the collaborative processing and data storage of bird data. The model provides secure storage and secure access control of bird data in the cloud, thereby ensuring the decentralized and secure storage of wild-bird data for multiple research institutions.
- 3 September 2020
- Published by Springer Nature in Multimedia Tools and Applications
- Vol. 80 (1) , 499-516
Fast image blending for high-quality panoramic images on mobile phones
- 3 September 2020
- Published by Springer Nature in Multimedia Tools and Applications
- Vol. 80 (1) , 499-516
Fast image blending for high-quality panoramic images on mobile phones
- 31 October 2019
- Published by Springer Nature
- p. 325-336
YNBIRDS: A System for Fine-Grained Bird Image Recognition
- 31 October 2019
- Published by Springer Nature
- p. 325-336
YNBIRDS: A System for Fine-Grained Bird Image Recognition