Distinguishing One Year and Two Year Old Canes of Red Raspberry Plant using Spectral Reflectance

摘要

Red raspberry (Rubus idaeus) is one of the important horticultural crops around the world. Various canopy management activities such as cane pruning, bundling and tying are used in this crop to improve light distribution and air-flow through plant canopies, which will reduce pest stress and improve crop yield and quality. This operation, however, is highly labor intensive, which threatens the long-term, sustainability of the red raspberry industry as labor availability is dwindling and labor cost is increasing rapidly. Mechanized or automated pruning and bundling system could be a key to ease this problem and increase returns to the growers. First step in developing an automated pruning system is to distinguish one-year-old canes (calledprimocanes) and two-year-old canes (calledfloricanes), which could then be used by a robotic system to selectively remove floricanes from the mix of primocanes and floricanes. Floricanes and primocanes look similar in terms of color, shape and size of the canes during dormant season. Hence, a non-imaging spectroscopy method was investigated in this study to utilize spectral signature of primocanes and floricanes, which can vary between two types of canes based on their difference in moisture content and chlorophyll concentration. Forty samples of each floricanes and primocanes were collected in Nov 2017 from a plot of‘Wakefield’ cultivar. Optimal wavebands were selected using Principal Component Analysis (PCA) and one-way ANOVA. Wavebands with the significance level of 5% were used. With a group of wavebands in the visible spectrum (596nm, 65nm, 676 nm and 716nm), primocanes and floricanes were distinguished with an accuracy of 91.7% using linear support vector machine. With a combination of wavebands from chlorophyll absorption and water absorption region (716nm, 856nm, 996nm, 1056nm, and 1396nm), a classification accuracy of 100% was achieved. Results show a promise for developing a multispectral sensor (with a few selected bands) for distinguishing between floricane and primocane. To our knowledge, this work represents the first study to compare the reflectance spectra signature to distinguish primocanes and floricanes of red raspberry plants.