Assessment of computer vision methods for motion tracking of planar mechanisms

Abstract

One of the main challenges on the use of planar mechanisms is to verify and monitor that the trajectories described by the mechanism correspond to those originally required. However, very few research studies have focused on tracking and monitoring the motion of target points located on the mechanisms during operation conditions. In this paper, a comparative study to evaluate the performance of several computer vision methods (CVMs) when used in motion tracking of planar mechanisms is presented. The aim is to compare and identify the best CVM, in terms of precision, speed, low cost, and computational performance, to track the movement of planar mechanisms. For this purpose, a case study corresponding to a planar four-bar mechanism is selected and analysed. The results show that the vision methods based on the homogeneous and non-homogeneous solution of the camera calibration matrix are a technological alternative for monitoring motion trajectories of planar mechanisms.

Leveraging Attention for Improved Discriminative Correlation Filters in Single Object Tracking

“…In a comprehensive comparison, the proposed Att-DCF approach is evaluated against 9 recent top trackers: LADCF [14], Eeco [32], BACF [33], C-cot [34], SRDCF [20], STA-PLE [23], DSST [18], SAMF [35], and KCF [17]. The evaluation is adopted using the One Pass Evaluation (OPE) [36], which uses distance precision and overlaps metrics to measure the trackers' accuracy. Using both distance precision and overlap metrics for OPE, the results show that the proposed tracker outperforms competing trackers in terms of both metrics as shown in Table 1, and Fig.…”