Detecting and Measuring Roots in 
Minirhizotron Images

Guang Zeng (Electrical and Computer Engineering)
Stan Birchfield (Electrical and Computer Engineering)
Christina E. Wells (Horticulture)

Minirhizotrons are small transparent tubes buried at an angle in the ground to observe the underground roots of neighboring plants. At regular intervals, researchers collect data by sliding a miniature camera into each tube and collecting still images of the roots seen through the tubes.  The result is an overwhelming amount of data that must be analyzed, traditionally by hand. The goal of this research is to automate the procedure of extracting, identifying, and measuring roots in minirhizotron images. We have developed an algorithm that applies two-dimensional matching filtering followed by local entropy thresholding to produce binarized images, from which roots are detected.   After applying a root classifier to discriminate fine roots from unwanted background objects, a root labeling method is implemented to identify each root in the image. Once a root is identified, its length and diameter are measured using Dijkstra's algorithm for obtaining the central curve and using the Kimura-Kikuchi-Yamasaki method for measuring the length of the digitized path.

Algorithm

Images with roots
		Correct
		Correct
		Correct
		Correct
		Correct
		Correct
		Correct
		Correct
		False Positive
		False Negative

Images without roots
		Correct
		Correct
		Correct
		Correct
		Correct
		Correct
		Correct
		Correct
		False Positive
		False Positive

Databases

There are two image databases available to download:

• Database1:  For developing and testing algorithms for root detecting and measuring (first reference).  Includes 100 non-root images and 150 images containing single or multiple young peach (Prunus persica) roots.  The detection / measuring results of our algorithm on this database are available here.
• Database2:  For training and testing algorithms for root discrimination (paper in preparation).  Includes 840 images from three types of plants: peach (Prunus persica), freeman maple (Acer x freemanii) and sweetbay magnolia (Magnolia virginiana). Also includes the discrimination results of our algorithm.

Note:  We have combined these databases into a single location, which also includes some detection results from our new algorithm.

Publications

• G. Zeng, S. T. Birchfield, and C. E. Wells.  Detecting and Measuring Fine Roots in Minirhizotron Images Using Matched Filtering and Local Entropy Thresholding.  Machine Vision and Applications, 17(4):265-278, 2006.

• G. Zeng.  Automatic Minirhizotron Root Image Analysis Using Two-Dimensional Matched Filtering And Local Entropy Thresholding.  Master's thesis, Dept. of Electrical and Computer Engineering, Clemson University, May 2005. (slides)