|POTHULA, ANAND - North Dakota State University
|IGATHINATHANE, C - North Dakota State University
Submitted to: Industrial Crops and Products
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 3/15/2015
Publication Date: 4/1/2015
Publication URL: http://handle.nal.usda.gov/10113/61318
Citation: Pothula, A.K., Igathinathane, C., Kronberg, S.L. 2015. Profile based image analysis for identification of chopped biomass stem nodes and internodes. Industrial Crops and Products. 70:374-382.
Interpretive Summary: Chemical fractions in the stems of crops such as corn and switchgrass have potential to be profitably converted to biofuels and bio-products if stem material from the crops can be converted in a cost effective manner. Stem nodes and internodes have different chemical composition and therefore could be used for different purposes. So, it is critical that effective means be developed to identify large quantities of stem nodes and internodes so that they can be sorted. Profile based image analysis has potential to quickly and accurately identify pieces of stem that are predominantly nodes or internodes. We developed a program to use within ImageJ software, which uses a relatively simple method based on visual outlines, to accurately identify nodes and internodes of switchgrass stems. The method can be easily applied to other chopped biomass and similar materials for profile based identification during biomass processing.
Technical Abstract: Because of their significant variation in chemical composition, segregation of chopped biomass into nodes and internodes helps in efficient utilization of these feedstocks. Stem internodes having low ash content are a better feedstock for bioenergy and biofuel applications than nodes. However, separation of these components is difficult because their characteristics are similar. In this study, we used the object profile identified differences for the node and internode components of scanned images of switchgrass and corn stalks. An algorithm was developed for digitally chopping the ends of the objects to facilitate node and internode identification, which would otherwise complicate the process especially dealing with cracked ends and projecting fibers. We developed an ImageJ plugin for the four image processing methods considered including rectangularity, solidity, width-, and slope-variation. Among the methods tested, width-variation gave the best identification accuracy (96.9% - 98.3%), followed by rectangularity (92.6% - 95.9%), solidity (86.3% - 90.7%), and slope-variation (68.9% - 82.2%). Rectangularity, a relatively simple method, and solidity, a standard ImageJ output, can be directly used to perform node and internode identification. The method can be easily applied to other chopped biomass and similar materials for profile-based identification.