Article Preview
Top1 Introduction
Nowadays, augmented reality (AR) has coexisted the real world with virtual objects. The technology has increased human experience in a virtual-reality intertwined world. It has grown in popularity over the last ten years, moving from laboratories into various real-life scenes (Van Krevelen & Poelman, 2010). However, there are numerous issues in AR, leading to the rise of innovations (Masood & Egger, 2019). Outdoor AR systems face challenges such as handling complex large-scale scenes, dealing with real-time occlusion, and rendering virtual objects. Although image segmentation could address the issues, it requires collaboration with other advanced techniques (Roxas et al., 2018). The collaboration is critical in the future trend of AR applications. Since traditional methods for sharing accurate spatial information are insufficient, various machine learning algorithms have been proposed to achieve low-cost and high-efficiency future collaboration systems (Zou et al., 2019). The volume of mobile and industrial AR applications is growing at an exponential rate; however, previous high latency, low precision, and unfriendly user experience have hampered the widespread adoption of AR systems (Huang et al., 2013). To overcome these limitations, registration (Hoff et al., 1996), tracking (Runz et al., 2018), image segmentation (Kirillov et al., 2019) and occlusion (Tang et al., 2020) approaches have been proposed with machine learning-improved visualization techniques, i.e., visual SLAM (vSLAM) and image segmentation. This paper provides an overview of the main obstacles of the AR visualizations and their potential solutions. The most recent computer vision technologies are concentrated, particularly machine learning-enhanced innovative technologies.
To date, there exists a lack of comprehensive literature review on the topic of applying the latest machine learning-improved computer vision in AR systems. However, there are some literature reviews related to other research communities in the AR systems, which are shown in Table 1.
As part of the digital transformation of industry, AR improves industrial efficiency, safety, compliance, and costs. Ling et al. (2017) discussed commercial trends in industrial AR. Palmarini et al. (2018) conducted a systematic literature review to identify the most relevant industrial AR technical limitations. Li et al. (2018) examined various Virtual Reality (VR) / AR prototypes, products, and training evaluation paradigms. Gattullo et al. (2020) systematically reviewed the literature on visual assets. Egger et al. (2020) investigated the current challenges and future directions of AR manufacturing. Costa et al. (2022) provided an overview of the current state of the art in AR human-robot collaboration and future development trends. Notably, user research on AR cybersecurity applications is severely lacking. Alzahrani et al. (2022) identified, described, and synthesized research findings on the cybersecurity of the AR industry.