Development and Validation of a Method for Measurement of Root Length in 2D Images

The plant root system analysis using traditional methods is complex, time-consuming, and in most cases does not provide the required accuracy. Therefore, new methods using digital image processing has been proposed. This paper presents a new algorithm for estimate the length of washed roots using digital image processing. First the algorithm loaded the image and extracted its resolution. Then the image was converted to grayscale and binarized. The objects from binary image were detected using its contours. The binary image was also used to thin the objects. After that the length was estimated based on resulting skeletons from previous procedure. The proposed algorithm was validated using copper wires. The copper wires length were measured using a caliper, and compared to estimated length by proposed algorithm and a method from literature, with different angles of inclination. When comparing estimated lengths from proposed algorithm and literature method, the proposed algorithm got the best results in 67,03% of test cases. Since plant roots are normally randomly arranged in the images, it is the greatest importance to develop a method of measuring root length that is invariant or has low sensitivity to rotation, and which is still precise and accurate.