研究目的
to explore the practicability of deep learning techniques for rice recognition in complex landscape regions, where rice is easily confused with the surroundings, by using mid-resolution remote sensing images.
研究成果
The proposed CNN model performed better than SVM in terms of user’s accuracy, producer’s accuracy, and overall accuracy. This innovative approach effectively facilitated rice identi?cation on mid-high resolution remote sensing images of complex landscape areas, which was not only broad the ways of rice information extraction but also meaningful in predicting grain yield, mitigating climate change and managing resources.
研究不足
The VI time series exerts great in?uence on the classi?cation model. Enough data sources should be gathered for constructing a ?tting time series curve, but uncontrollable clouds and rain coverage lead to lacking data collecting in southern China, particularly in growing seasons. Model training is also signi?cant for error analysis, involving quantity of samples, layers of network, iterations, etc. The landscape is complex and diverse in hilly areas, and some misclassi?cation may occur.
1：Experimental Design and Method Selection
A novel pixel-level classi?cation approach that uses convolutional neural network (CNN) model to extract the features of enhanced vegetation index (EVI) time series curve for classi?cation. A transfer learning strategy is utilized to ?ne tune a pre-trained CNN model and obtain the temporal features of the EVI curve. Support vector machine (SVM), a traditional machine learning approach, is also implemented in the experiment.
2：Sample Selection and Data Sources
Eight HJ-1 A/B images from 13 May to 3 November 2017 were collected to generate EVI datasets for machine learning and classi?cation. The study area is Zhuzhou City, situated in the east of Hunan Province, China.
3：List of Experimental Equipment and Materials
HJ-1 A/B satellite data, ENVI5.2 image processing software, FLAASH module for atmospheric correction, GPS device for ground-truth data collection.
4：Experimental Procedures and Operational Workflow
1. Calculate the VI of multitemporal satellite data and reconstruct the time series curve as input. 2. Develop a deep learning model for classi?cation based on the framework of the Convolutional Architecture for Fast Feature Embedding, introducing the strategy of transfer learning. 3. Conduct model training and validation by selecting different samples. 4. Use SVM and deep learning model for classi?cation for contrast.
5：Data Analysis Methods
Evaluate the accuracy of the two models (CNN and SVM) using ground truth points and visual identi?cation information from Google Earth.
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
