| Title |
Stereo Vision for Real-time Measurement of Driving Vehicle Dimensions and a Performance Comparison with Commercial LiDAR |
| Authors |
김창일(Chang-il Kim) ; 박진욱(Jinuk Park) ; 임용석(Yong-seok Lim) ; 박용주(Yongju Park) |
| DOI |
https://doi.org/10.5573/ieie.2026.63.1.53 |
| Keywords |
Stereo vision system; Vehicle dimension measurement; Real-time stitching; Falling objects |
| Abstract |
Accidents caused by falling objects due to improper loading of trucks seriously threaten the safety of vehicles in operation and cause damage to roads. Especially, the risk of falling objects is high when cargo is loaded in excess of the specified dimensions. Therefore, research is actively being conducted to crack down on improper loading based on technology that measures vehicle specifications. While commercial technologies such as laser scanners and LiDAR exist, their widespread adoption is challenging due to high initial deployment costs and complex maintenance. To compensate for these shortcomings, research is currently being conducted to analyze the shape of a vehicle using images acquired from multiple cameras. However, deriving real-time analysis results for moving vehicles has limitations. To overcome this, this paper proposes a stereo vision system using two cameras to analyze the shape and determine the dimensional specifications of moving vehicles in real-time. To achieve this, we implemented a function to verify vehicle entry by using real-time background updates, and integrated the segmented vehicle images by applying a improved algorithm that utilizes the bilateral filter, morphological operations, and a matching score. Performance validation, conducted by introducing a commercial LiDAR system, showed that the proposed system exhibited comparable or superior performance, with measurement accuracies of 97.33% for height and 96.28% for length. Especially, the system demonstrated superior performance in width measurement accuracy at 97.63% (6.72%p increase) compared to commercial LiDAR. Furthermore, it is three times more affordable than LiDAR, which provides a significant competitive edge. |