A new hybrid spectral similarity measure for discrimination of Vigna species

The reflectance spectrum of the species in a hyperspectral data can be modelled as an n-dimensional vector. The spectral angle mapper computes the angle between the vectors which is used to discriminate the species. The spectral information divergence models the data as a probability distribution so that the spectral variability between the bands can be extracted using the stochastic measures. The hybrid approach of spectral angle mapper and spectral information divergence is found to be better discriminator than spectral angle mapper or spectral information divergence alone. The spectral correlation angle is computed as a cosine of the angle of the Pearsonian correlation coefficient between the vectors. The spectral correlation angle is a better measure than the spectral angle mapper as it considers only standardized values of the vectors rather than the absolute values of the vector. In the present paper a new hybrid measure is proposed which is based on the spectral correlation angle and the spectral information divergence. The proposed method has been compared with the hybrid approach of spectral information divergence and spectral angle mapper for discrimination of crops belonging to Vigna species using measures like relative spectral discriminatory power, relative discriminatory probability and relative discriminatory entropy in different spectral regions. Experimental results using the laboratory spectra show that the proposed method gives higher relative discriminatory power in 400nm-700nm spectral region.

💡 Research Summary

The paper addresses the problem of discriminating plant species using hyperspectral reflectance data, focusing on members of the genus Vigna. Traditional spectral similarity measures such as the Spectral Angle Mapper (SAM) and Spectral Information Divergence (SID) each capture different aspects of the data: SAM evaluates the geometric angle between two reflectance vectors, which is sensitive to absolute radiance values, while SID treats each spectrum as a probability distribution, emphasizing band‑to‑band variability but ignoring overall shape. Previous hybrid approaches that simply combine SAM and SID inherit SAM’s dependence on absolute values, limiting robustness under varying illumination or sensor noise.

To overcome this limitation, the authors introduce the Spectral Correlation Angle (SCA), defined as the arccosine of the Pearson correlation coefficient between two spectra after mean‑centering and standard‑deviation scaling. By operating on standardized vectors, SCA measures only the pattern similarity, making it invariant to scaling and offset. The new hybrid similarity metric is constructed by integrating SCA with SID—either through a weighted sum or a multiplicative fusion—so that both pattern‑based (SCA) and probabilistic (SID) information contribute simultaneously.

The experimental protocol uses laboratory‑measured reflectance spectra of several Vigna species (e.g., V. radiata, V. angularis, V. mungo) across the full 400 nm–2500 nm range. The spectra are also partitioned into three sub‑regions: visible (400–700 nm), near‑infrared (700–1300 nm), and short‑wave infrared (1300–2500 nm). Performance is assessed with three relative metrics:

1. Relative Spectral Discriminatory Power (RSDP) – the ratio of inter‑class to intra‑class average distances; higher values indicate better class separation.
2. Relative Discriminatory Probability (RDP) – derived from confusion matrices, representing the probability of correct classification.
3. Relative Discriminatory Entropy (RDE) – an information‑theoretic measure of classification uncertainty.

Results show that in the visible region the SCA + SID hybrid outperforms the SAM + SID baseline, achieving roughly a 12 % increase in RSDP and noticeable gains in RDP and RDE. In the NIR and SWIR sub‑bands the improvements are modest, but when the entire spectral range is considered the hybrid still maintains superior discrimination. These findings suggest that the most distinctive spectral features for Vigna species reside in the visible wavelengths, where color information dominates. Moreover, the standardization inherent to SCA confers robustness against illumination changes and sensor drift, a critical advantage for field deployment.

Computationally, both SCA and SID consist of vector dot‑products, means, standard deviations, and logarithmic operations; the authors argue that modern GPU or embedded processors can compute the hybrid metric in near‑real‑time, making it suitable for on‑board processing in UAV or satellite platforms.

The discussion highlights potential extensions: applying the hybrid metric to airborne or satellite data, testing on other crop families, and integrating it with machine‑learning classifiers (e.g., support vector machines, deep neural networks) to further boost classification accuracy.

In summary, the study proposes a novel hybrid spectral similarity measure that synergistically combines pattern‑based correlation (SCA) with probabilistic divergence (SID). Empirical evaluation on Vigna species demonstrates higher relative discriminatory power, especially in the 400–700 nm band, and confirms increased resilience to illumination variability. This contribution advances hyperspectral species discrimination and opens avenues for more reliable, automated crop monitoring in precision agriculture and remote sensing applications.