关键词: 光谱特征尺度, ASD, 水稻重金属胁迫, 小波信息熵, 小波分形

Abstract: In order to explore the sensitivity of hyperspectral remote sensing monitoring to heavy metal stress in rice at different spectral scales based on ASD hyperspectral data under different pollution levels, 450~900 nm of ASD spectral band was selected as the sensitive band of heavy metal stress, and the multi-scale wavelet coefficients generated by DB5 wavelet transform were used to simulate different spectral resolutions. Furthermore, the wavelet fractal dimension and wavelet detail coefficient entropy were determined to identify the turning point of rice spectral characteristic scale，the reliability and accuracy of the identified characteristic scales were verified by constructing stress-related chlorophyll spectral indices MCARI/OSAVI, NDSI_R and Depth. The results indicated that: ① The wavelet information entropy calculated by the detail coefficients of each scale of wavelet decomposition had obvious characteristic turning points of distinguishing stress levels near the 5～7 scale of decomposition. ②The statistical analysis of wavelet fractal dimension indicated that the  scale 5  was the most obvious turning point of the spectral curve scale of rice under different scales of heavy metal stress and reconstruction. At scale 5, the peak and valley details of the spectral curve were better reflected. ③ The study of the spectral characteristics of heavy metal pollution in rice not only retained the main characteristic information of spectral information, but also minimized the amount of spectral data, which improved the efficiency of monitoring heavy metal pollution in rice,and provided a new means for environmental monitoring.

Key words: characteristic spectral scale, ASD, heavy metal stress, wavelet information entropy, wavelet fractal