摘要：AbstractReal-time monitoring of crop growth parameters in plant factory can provide useful information about accurate assessment of their growth status for precision crop management. Plant weight is one of the most important biophysical properties used to determine the optimum time for harvesting. Conventional plant weight measurements are destructive and laborious. An on-the-go image processing system consisting of image acquisition and weight estimation was developed to generate a fresh weight map of plants grown in hydroponic solutions. Key technologies developed in this study are real-time image processing and spatial mapping methods that estimate the fresh weights of individual lettuces. Images were automatically captured with a low cost web camera and processed using a MYRIO-based embedded controller. The camera and embedded system moved along an XY axis frame above a plant growing bed (0.94 x 1.8 m) using two stepping motors and linear actuators. The image preprocessing algorithm consisted of two main subroutines, i.e., image segmentation and target plant recognition. For the image segmentation, the S channel of the HSV color space and Otsu’s threshold were used to separate the plants from the background. The target plant was identified based on location information of the growing bed holes in captured images. The plant weight was estimated using calibration equations previously developed that relate the pixel numbers of lettuce images to their actual fresh weights in conjunction with the use of a two-point normalization method. The accuracy of the fresh weight determined by the developed embedded system was confirmed by a highly linear relationship with a slope near 1.0 and a coefficient of determination (R2) of 0.95 with a processing time of within 4 s. In addition, it was possible to generate a spatial map of the fresh weights of lettuces grown in a cultivation bed, which could be used to estimate their yields prior to harvesting.
