Comparative Analysis of Deep Neural Networks for the Detection and Decoding of Data Matrix Landmarks in Cluttered Indoor Environments

Abstract

Data Matrix patterns imprinted as passive visual landmarks have shown to be a valid solution for the self-localization of Automated Guided Vehicles (AGVs) in shop floors. However, existing Data Matrix decoding applications take a long time to detect and segment the markers in the input image. Therefore, this paper proposes a pipeline where the detector is based on a real-time Deep Learning network and the decoder is a conventional method, i.e. the implementation in libdmtx. To do so, several types of Deep Neural Networks (DNNs) for object detection were studied, trained, compared, and assessed. The architectures range from region proposals (Faster R-CNN) to single-shot methods (SSD and YOLO). This study focused on performance and processing time to select the best Deep Learning (DL) model to carry out the detection of the visual markers. Additionally, a specific data set was created to evaluate those networks. This test set includes demanding situations, such as high illumination gradients in the same scene and Data Matrix markers positioned in skewed planes. The proposed approach outperformed the best known and most used Data Matrix decoder available in libraries like libdmtx.

A Data Matrix Code Recognition Method Based on L-Shaped Dashed Edge Localization Using Central Prior

“…Moreover, accurate and efficient localization can reduce the complexity and time consumption of decoding. Therefore, the majority of researchers [7][8][9][10][11]35,36] prioritize efforts to enhance the accuracy and efficiency of the localization to improve overall DM code recognition performance.…”

“…However, the detection is time-consuming because of the large number of parameters in Faster R-CNN. To solve this issue, Almeida et al [36] conducted a comprehensive analysis and compared experiments on representative deep learning-based object detection methods [18,19,21,25] with different backbones [39][40][41][44][45][46], and they proposed to use YOLOV4 [21] as the detector to acquire balanced detection performance between accuracy and latency. To deal with the problem that images captured by mobile cameras are usually of low quality with poor contrast, a deep learning-based method for industrial DM code was proposed by Liao et al [47] to learn the colors of two adjacent modules of a DM symbol.…”

“…For these interference issues, unlike existing methods [7][8][9][10][11][12][13] that rely on the L-shaped solid edge for DM code localization, our method relies on the L-shaped dashed edge with stronger interference-resisting capabilities for DM code localization. Existing methods [35,36] yield results that still contain the interference issues caused by protrusions and interruptions on the L-shaped solid and dashed edges when simply utilizing object detection methods based on deep learning network. In contrast, our approach leverages an object detection method to yield the center of the DM code, which is then used to build a two-level screening strategy combining general constraints and central constraints for locating the L-shaped dashed edge.…”

Using Different Types of Artificial Neural Networks to Classify 2D Matrix Codes and Their Rotations—A Comparative Study

“…In the first phase of their research, Hansen et al [ 6 ] utilized the YOLO object detection algorithm (based on the Darknet-19 CNN architecture) to detect 1D and QR codes in a whole image, while in the second phase, another angle prediction network (also based on Darknet-19) was used. Almeida et al [ 7 ] investigated different types of object detectors (Faster R-CNN, SSD, YOLO) based on CNNs to locate Data Matrix codes used as navigation landmarks. YOLOv4 was found to be the best detector, followed by a conventional decoder (libdmtx).…”

DM Code Key Point Detection Algorithm Based on CenterNet

“…Wan et al [20] proposed a lightweight CenterNet network for multi-scale 2D code positioning by lightweighting the backbone of the original CenterNet network to Cross Stage Partial Network Tiny (CSPDarkNet53-Tiny), while adding SPP modules to the network structure for multi-scale fusion and replacing the normal convolution in the detection head part with a depth-separable convolution, which realizes fast 2D code recognition in low configuration conditions. Almeida et al [21] compared and analyzed several deep neural networks for detecting and decoding DM code in complex indoor environments. They investigated various deep neural network architectures ranging from two-stage to single-stage and evaluated their performance detecting DM code in indoor environments.…”