Prediction model of plant leaf wilting using 3-D machine vision

Wilting of plants appears when the water supply of plants is insufficient. Quantitative identification of wilting phenomena of plant stems, leaves and other parts is of important practical significance to improve agricultural production and efficient water irrigation. A scanning device based on the principle of laser ranging oblique is used to obtain real-time 3-D plant images. Then leaf curl statistical index, fractal dimension and the DC component of two-dimensional Fourier spectrum were extracted as wilting index to quantify the degree of wilting plants. Three kinds of wilting indexes were tested on zucchini, gourds, pumpkins and okra leaves to find out the correlations with wilting extent. The experiments show that each wilting index had a good correlation with wilting extent (0.82 or more are reached). On this basis, a multiple linear regression model of three kinds of wilting indexes, vapor pressure deficit (VPD) and photosynthetically active radiation (PAR) was built.