Estimation of the Leaf Area Index, Leaf Fresh Weight, and Leaf Length of Chinese Chive (Allium tuberosum) Using Nadir-looking Photography in Combination with Allometric Relationships

Abstract

In horticultural leafy vegetable production, continuously monitoring crop size indicators such as the leaf area index (LAI), leaf fresh weight (LFW), and leaf length (LL) is of practical value because these indicators are related to crop yields and harvest timing. The aim of this study was to develop a method that enables the continuous, automatic estimation of the LAI, LFW, and LL of a Chinese chive (Allium tuberosum) canopy by combining timelapse photography with allometric equations. LAI was estimated based on the gap fractions of nadir photographs (i.e., the fractions of nonleaf area), which were retrieved using the deep learning framework DeepLabv3+ with satisfactory accuracy (mean intersection over union, 0.71). This photographically estimated LAI (LAIphoto) corresponded well with the destructively measured LAI (LAIdest) (LAIphoto = 0.96LAIdest, R2 = 0.87). LAIphoto was then used as the input of allometric regression equations relating LAIphoto with LFW and LL. A power function (y = axb) fit the observed LAIphoto–LFW and LAIphoto–LL relationships well (R2 = 0.89 and 0.74, respectively). By combining nadir timelapse photography with the allometric equations, changes in the LFW and LL of a Chinese chive canopy were estimated successfully for a 9-month cultivation period. Our approach can replace time-consuming, labor-intensive manual measurements of these crop size indicators for Chinese chive and may be applicable to other crops with different parameter sets.

Keywords

• beer–lambert law
• gap fraction
• image analysis
• leaf area index
• semantic segmentation
• timelapse digital camera