Nondestructive measurement method for greenhouse cucumber parameters based on machine vision

The use of machine vision technology for nondestructive online measurements of cucumber parameters was investigated. This technology was first used to capture images of a cucumber canopy. Next, a segmentation algorithm (excess green minus excess red (ExG-ExR)) was used to extract the cucumber canopy area and image parameters (i.e., coverage ratio, canopy length and canopy width). These parameters were combined with those obtained by manual measurements (i.e., stem height, stem diameter, leaf number, and fruit number) to generate five inversion models for four cucumber growth parameters. The results showed that the ExG-ExR segmentation method yielded a 99.5% contact ratio and a 98.2% recognition rate in the extraction of the cucumber canopy region. The inversion models were validated with new images using the following three different cultivation modes: 4 × 2, 4 × 3 and 4 × 4. The inversion results showed that the coefficients of determination (R2) between the measured values and inversion values of stem height, stem diameter, leaf number, and fruit number exceeded 0.921, 0.899, 0.95 and 0.908, respectively. Thus, the inversion method can provide nondestructive online measurements of cucumber parameters. © 2015 Asian Agricultural and Biological Engineering Association.