Development of low-cost leaf spectroscopy based on featured wavelength extraction by fix-bandwidth Gaussian mixture model

Light spectrum analysis of leaves plays an essential role in measuring the contents of carbohydrates and other important compound like proteins. And it is also useful in evaluating the status of vegetation by remote sensing. Here, an approximation method considering the optical response characteristics of multispectral silicon (Si) sensor for the reflection or absorption spectra of leaves is proposed. The light spectrum is analysed by Gaussian mixture models (GMM) with fixed bandwidth firstly, and then the optimal model parameters are derived and validated to balance the complexity of Si sensor and the 2-norm deviation of spectra, and next a low-cost framework for the leaf spectroscopy with wireless node is illustrated according to the specific centre-wavelengths and bandwidths of the GMM, and finally, an experimental prototype with 18-channel Si sensor node and the developed mobile APP is demonstrated. The simplified strategy and its realization for leaf spectroscopy cast light on large-scale applications in future agriculture.