研究目的
To determine crop types using hyperspectral images collected by a ground scanner through spectral mixture analysis and to evaluate the effectiveness of different endmember extraction algorithms for precise agriculture.
研究成果
The study demonstrated the effectiveness of spectral mixture analysis for crop type determination using hyperspectral images. SISAL showed high accuracy for green/red crop determination, while VCA was most effective for discriminating lettuce and perilla leaf. Future work will focus on developing optimized abundance map estimation techniques for higher accuracy.
研究不足
The study includes limited kinds of crops and image size, necessitating evaluation on more hyperspectral data to analyze the performance of the algorithms.
1：Experimental Design and Method Selection
The study utilized spectral mixture analysis for hyperspectral image processing, employing N-FINDR, VCA, and SISAL algorithms for endmember extraction and FCLSU for classification.
2：Sample Selection and Data Sources
Hyperspectral image data at canopy level was collected by a SPECIM hyperspectral scanner (PS-FW-11-V10E), focusing on green crops.
3：List of Experimental Equipment and Materials
SPECIM hyperspectral scanner (PS-FW-11-V10E), dark reference, 18% white reference.
4：Experimental Procedures and Operational Workflow
Radiometric calibration was performed, noise bands were removed, and the data was subset to 100×100 pixels. Crop type determination was performed in two cases based on vegetation color and crop species.
5：Data Analysis Methods
Endmember extraction was performed using N-FINDR, VCA, and SISAL algorithms. Classification was based on abundance maps estimated by FCLSU, with results compared to manually collected endmembers.
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
