Performant barcode decoding for herbarium specimen images using vector-assisted region proposals (VARP)

Abstract

Premise: The scale and associated costs of herbarium digitization make process automation appealing. One such process for many workflows is the association of specimen image files with barcode values stored with the specimen. Here, an innovation is presented that improves the speed and accuracy of decoding barcodes from specimen images. Methods and Results: Geometric features common in barcodes are used to identify the regions of specimen images that are likely to contain a barcode. The proposed regions are then combined into a significantly reduced composite image that is decoded using traditional barcode reading libraries. Tested against existing solutions, this method demonstrated the highest success rate (96.5%) and the second fastest processing time (617 ms). Conclusions: This method was developed to support a larger effort to automate specimen image post-processing in real-time, highlighting the importance of execution time. Although initially designed for herbarium digitization, this method may be useful for other high-resolution applications.