Machine vision for vial positioning detection toward the safe automation of material synthesis

Leslie Ching Ow Tiong,Hyuk Jun Yoo,Na Yeon Kim,Kwan-Young Lee,Sang Soo Han,Donghun Kim

DOI: https://doi.org/10.48550/arXiv.2206.07272

2022-06-15

Abstract:Although robot-based automation in chemistry laboratories can accelerate the material development process, surveillance-free environments may lead to dangerous accidents primarily due to machine control errors. Object detection techniques can play vital roles in addressing these safety issues; however, state-of-the-art detectors, including single-shot detector (SSD) models, suffer from insufficient accuracy in environments involving complex and noisy scenes. With the aim of improving safety in a surveillance-free laboratory, we report a novel deep learning (DL)-based object detector, namely, DenseSSD. For the foremost and frequent problem of detecting vial positions, DenseSSD achieved a mean average precision (mAP) over 95% based on a complex dataset involving both empty and solution-filled vials, greatly exceeding those of conventional detectors; such high precision is critical to minimizing failure-induced accidents. Additionally, DenseSSD was observed to be highly insensitive to the environmental changes, maintaining its high precision under the variations of solution colors or testing view angles. The robustness of DenseSSD would allow the utilized equipment settings to be more flexible. This work demonstrates that DenseSSD is useful for enhancing safety in an automated material synthesis environment, and it can be extended to various applications where high detection accuracy and speed are both needed.

Computer Vision and Pattern Recognition,Artificial Intelligence

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to improve the accuracy of reagent bottle position detection through machine vision technology in an automated material synthesis laboratory, so as to enhance the safety of experiments. Specifically, the paper focuses on dangerous accidents that may be caused by robotic automated operations in the absence of manual monitoring, especially those caused by machine control errors. These accidents may occur when the positions of reagent bottles (such as vials, beakers, test tubes, etc.) are incorrect. For example, if the reagent bottle is not placed correctly during the solution stirring process, an accident may occur.

To meet this challenge, the paper proposes a new object detector based on deep learning - DenseSSD (Densely Connected Single - Shot Detector), aiming to achieve high - precision reagent bottle position detection in complex and variable environments. DenseSSD improves the learning of feature representation by introducing a dense connection mechanism, thereby achieving an average precision (mAP) of over 95% in detecting both empty reagent bottles and reagent bottles filled with solutions, significantly exceeding the existing YOLO and SSD models. In addition, DenseSSD has high robustness to environmental changes (such as changes in solution color or test perspective), which makes the device setup more flexible in the laboratory environment. Finally, the paper also designs a safety alarm module for remotely and immediately notifying researchers when a failure case is detected, further improving the safety of the laboratory.