研究目的
Investigating the classification of multi-source remote sensing data using fully convolutional networks and post-classification processing to improve accuracy.
研究成果
The proposed Fusion-FCN methodology for multi-source remote sensing data classification achieved an overall accuracy of 80.78%, ranking first in the 2018 IEEE GRSS Data Fusion Contest. Future work will focus on automatically learning suitable weighting factors for data fusion.
研究不足
The method performs poorly on classes like water, crosswalks, and unpaved parking lots. Future work will investigate this phenomenon.
1：Experimental Design and Method Selection:
The study employs a new data fusion methodology named Fusion-FCN for classifying multi-source remote sensing data, utilizing fully convolutional networks (FCNs) and post-classification processing.
2：Sample Selection and Data Sources:
Three types of data are used: LiDAR data, hyperspectral images, and very high resolution images.
3：List of Experimental Equipment and Materials:
Not explicitly mentioned.
4：Experimental Procedures and Operational Workflow:
The methodology includes preprocessing of data, Fusion-FCN architecture for feature learning and fusion, and post-classification processing to correct misclassifications.
5：Data Analysis Methods:
The performance is evaluated based on overall accuracy and kappa coefficient.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
