Model-based in-situ measurement of pakchoi leaf area

Leaf area measurement is of great significance in plant growth process monitoring. It poses challenges to perform an unattended in-situ measurement, arising from quantifying the 3-dimensional pakchoi leaf surface. Conventional non-destructive measurement techniques, which usually take its projection on the horizon plane of the leaf area, inevitably cause considerable measurement errors. In order to improve the measurement precision for leaf area, the exemplar pakchoi leaf was modeled as a complete or a piecewise spatial plane to approximate the actual leaf surface, and a machine-vision based ad hoc measuring platform was developed to conduct the in-situ measurement. First, the leaf image was captured by a stereo vision system and segmented via a semi-automatic process to obtain its projective area and spatial inclination angle. Second, pakchoi leaves were modeled as spatial surfaces regarding to their projected counterparts. Third, leaf areas were calculated according to the established planar spatial model, acquired inclination angles and projective areas. The experimental comparison among the lattice-based monotype method, projection method, and the model-based method, whose results are denoted as MA, PA, and EA respectively, showed that the proposed framework could simultaneously meet the accuracy and non-destructive measurement requirements. The constructed platform also provided a cost-effective semi-automatic measurement approach for continuously in-situ monitoring of pakchoi growth during its whole cultivation period. It is further suggested from the experimental results that the proposed methodology can offer a generic measurement solution to various kinds of plant physiological and ecological studies in future researches.