Canopy detection beyond the field: Colored backgrounds impact precision of Canopeo

Abstract

The development of Canopeo as a close-range remote sensor for measuring ground cover fraction (GCF) offered farmers and scientists an accurate, simple, low-cost tool for monitoring health and development throughout the plant lifecycle. However, a significant obstacle to image-based monitoring of plant performance is the difficulty of object distinction between plant and background sharing similar colors. The overall goal of this research was to test Canopeo's sensitivity for detecting GCF when plants were imaged on different colored backgrounds in a greenhouse environment. We therefore tested Canopeo's ability to detect plant versus non-plant pixels in each image (resolution 72 × 72) using ten complex flat backgrounds. Multicolored backgrounds resembling flooring which may be found in a greenhouse setting (concrete, brick painted white, natural wood plank, and wood painted white with scuffs) resulted in least amount of deviation (20). Canopeo demonstrated greatest underestimation GCF for Viburnum sp. and E. pinnatum cv. on a red background. When GCF of the green background was omitted, the r2 value of 0.75, or goodness of fit, suggest approximately 75% of the sampling variation can be described by the background color and not experimental error. Canopeo is an easily accessible tool for researchers and farmers to monitor plant growth and development on a diversity of backgrounds beyond soil and field settings.

Keywords

• GCF
• Smartphone-based apps
• Mobile-applications
• Precision agriculture
• Image analysis
• Phenotyping