研究目的
Evaluating the effect of utilizing different types of signatures on plant detection success in hyperspectral aerial images.
研究成果
The study demonstrated that using a plant representative signature (PRS) for plant detection in hyperspectral aerial images yields higher detection performance compared to using separate organ signatures. The method is simple and useful, with potential applications in identifying different plant species.
研究不足
The study is limited to corn plant detection and the method's effectiveness on other plant species needs further investigation. The spectral signatures' variability due to physiological changes in plants over time was not addressed.
1：Experimental Design and Method Selection:
The study focused on the detection of corn plants from the air using spectral signatures of leaf, stem, and tassel separately and creating a plant representative signature (PRS) by averaging selected regions. SAM and GLRT algorithms were used for target detection.
2：Sample Selection and Data Sources:
Hyperspectral images taken from 10m distance to target plant were used to generate spectral signatures.
3：List of Experimental Equipment and Materials:
HySpex VNIR-1800 camera model was used to obtain hyperspectral images.
4：Experimental Procedures and Operational Workflow:
Pre-processing steps included radiometric and geometric correction, removal of atmospheric bands, and spectral normalization. Target detection was performed using SAM and GLRT algorithms.
5：Data Analysis Methods:
Performance evaluation was made by Receiver Operating Characteristic (ROC) curves.
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
