A computational method to assist evaluation of root length of soybean based on digital image processing

• Published in: Australian Journal of Crop Science Vol. 13; no. 11; pp. 1740 - 1748
• Main Authors: Name, Marcio Hosoya, Maruyama, Teruo Matos, Falate, Rosane, Fonseca, Adriel Ferreira da
• Format: Journal Article
• Language: English
• Published: Lismore, N.S.W Southern Cross Publishers 01.11.2019
• Subjects: Computational intelligence; Digital images; Growth; Nutrition; Plants; Root crops; Soybean
• ISSN: 1835-2693; 1835-2707
• DOI: 10.21475/ajcs.19.13.11.p1492

The evaluation of the root system is important for better understanding the effects of nutrient management on soil and plant nutrition. However, root system studies and culture are slow and show low accuracy. In this context, digital image processing may be an alterna-tive. The objective of this research was to develop a computational method to assist evaluation of the soybean root growth. Initially, the free and open access software, available at: http://rm.deinfo.uepg.br/, was developed in Java platform with the OpenCV library supply through the plug-in JavaCV. To evaluate the software, copper wires with 10 mm, 20 mm, and 50 mm of length manually meas-ured using callipers. They were scanned with a resolution of 300 dpi and then images were loaded in the software. Variation coeffi-cients between 0.01 and 2.99 % were obtained. Subsequently, the samples of soybean roots were scanned and the results of devel-oped software and Safira software were correlated with those from the line-intersect method. The determination coefficients (R = 0.999) of the developed software, on average, were better than those obtained with Safira software (R = 0.733), when comparing with the line-intersect method. Therefore, the proposed method was accurate for length measurements of soybean roots.