Geometric plant properties by relaxed stereo vision using simulated annealing

An important task in precision agriculture is to monitor the growth conditions of the crop plants. Compared to traditional monitoring techniques based on remote sensing from aircraft or satellite, ground-based computer vision techniques offer the advantage of allowing crop analysis on single plant scale. This article investigates the potential of using area-based binocular stereo vision for three-dimensional (3-D) analysis of single plants and estimation of geometric attributes such as height and total leaf area. During the stereo matching process, multiple candidates for pixel-to-pixel correspondences are found, and as a novel approach, the use of the simulated annealing (SA) method is proposed in order to find the best candidate, taking neighboring pixels into consideration. Under laboratory conditions, tests have been conducted on 10 young wheat plants. The results show that plant height and leaf area can be estimated by means of the proposed method. Using proper parameter settings, the use of simulated annealing improves the estimation considerably.