Estimation of soybean leaf area, edge, and defoliation using color image analysis

Abstract Soybeans are an extremely important crop in the United States, providing the largest source of animal protein feed and a main source of vegetable oil in the country. The goal of this research is to identify methodologies to estimate percent defoliation of the soybean canopy and leaves using RGB images taken in the field. The Mahalanobis distance classification method was used to process sets of images and calculate leaf area (number of pixels) corresponding to two classes (leaf and background) with eight different color groups. The Canny edge detection algorithm provided an efficient method for detecting leaf edges, and threshold t2 = 20 was found to be the optimal value for estimating soybean leaf edge. The segmentation results showed a performance of 96% for soybean leaves using Mahalanobis distance classification. Two statistical regression models (polynomial and logistic regression) for defoliation of soybean were developed based on individual images of trifoliate leaves taken from the field. The models both utilized leaf area and edge to provide estimates of soybean defoliation; however, a logistic equation has potential to provide greater understanding and more accurate estimates of defoliation with variations, especially at low defoliation. The R2 and root mean square error (RMSE) of estimated and observed defoliation of trifoliate leaves were 0.90 and 6.16%, respectively. The validation of soybean canopy defoliation and its corresponding trifoliate leaves defoliation also provided reasonable correlation (R2 = 0.96 and RMSE = 1.85%). This approach could lead to use of remotely sensed imagery for estimating defoliation in soybeans and timely intervention with integrated pest management strategies.

[1]  R. L. Jensen,et al.  A Method of Estimating Insect Defoliation of Soybean , 1977 .

[2]  M. Kogan,et al.  Sampling Methods in Soybean Entomology , 1980, Springer Series in Experimental Entomology.

[3]  László Györfi,et al.  A Probabilistic Theory of Pattern Recognition , 1996, Stochastic Modelling and Applied Probability.

[4]  D. Holshouser,et al.  Evaluation of the LAI-2000 Plant Canopy Analyzer to Estimate Leaf Area in Manually Defoliated Soybean , 2002 .

[5]  Feiping Nie,et al.  Learning a Mahalanobis distance metric for data clustering and classification , 2008, Pattern Recognit..

[6]  María-Paz Diago,et al.  Grapevine Yield and Leaf Area Estimation Using Supervised Classification Methodology on RGB Images Taken under Field Conditions , 2012, Sensors.

[7]  Wei Li,et al.  Combining discriminant analysis and neural networks for corn variety identification , 2010 .

[8]  Pravin Yannawar,et al.  Comparison of Edge Detection Methods in Leaf Shape Analysis For Plant Classification , 2015, CVPR 2015.

[9]  D Marr,et al.  Theory of edge detection , 1979, Proceedings of the Royal Society of London. Series B. Biological Sciences.

[10]  Sandeep Kumar.  LEAF COLOR, AREA AND EDGE FEATURES BASED APPROACH FOR IDENTIFICATION OF INDIAN MEDICINAL PLANTS , 2012 .

[11]  F. Hall,et al.  Feeding behavior of Plutella xylostella and Leptinotarsa decemlineata on leaves treated with Bacillus thuringiensis and esfenvalerate , 1993 .

[12]  W. Fehr,et al.  Stage of Development Descriptions for Soybeans, Glycine Max (L.) Merrill , 1971 .

[13]  John F. Canny,et al.  A Computational Approach to Edge Detection , 1986, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[14]  Douglas A. Landis,et al.  An Inexpensive, Accurate Method for Measuring Leaf Area and Defoliation Through Digital Image Analysis , 2002, Journal of economic entomology.

[15]  Rabia Masood,et al.  Plants Disease Segmentation using Image Processing , 2016 .

[16]  T. Mitchell-Olds,et al.  Induced plant defense responses against chewing insects. Ethylene signaling reduces resistance of Arabidopsis against Egyptian cotton worm but not diamondback moth. , 2000, Plant physiology.