Fine localization and distortion resistant detection of multi-class barcode in complex environments

Barcode, including one-dimensional (1D) barcode and two-dimensional (2D) barcode, can be seen almost anywhere in our lives. In many barcode-based mobile systems, different barcodes will appear simultaneously with different angles, shapes, and image quality. Barcode localization is a significant prerequisite for barcode decoding in these applications. In this paper, we propose a region-based end-to-end network to finely localize and classify 1D barcode and Quick Response (QR) code in complex environments. Two special layers are designed in our network. One is a quadrilateral regression layer to localize arbitrary quadrilateral bounding boxes, and another is a Multi-scale Spatial Pyramid Pooling (MSPP) layer to improve the detection accuracy of small-scale barcodes. Extensive experiments on existing public datasets and our own dataset have verified the effectiveness of proposed layers. We also demonstrate that our method can resist some distortions by simulating barcode images of different image qualities. What’s more, a human decoding experiment is also performed to prove the effectiveness of our method as a preprocessor for QR code decoding.