.png){kind=icon}
.png){kind=icon}
Return
Single-view object height estimation using ground point detection
.png){kind=icon}
PRE
AI
DOI：10.1111/1556-4029.70342.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Extracting reliable object height information from surveillance images is critical for understanding incidents and enhancing situational awareness. Such information supports key forensic tasks, including suspect tracking and extracting evidential object details from surveillance images. Object height estimation in conventional methods requires that both the reference object with known height and the target object be vertically aligned on the same plane. This constraint significantly limits their applicability in real-world forensic scenarios, where objects are often positioned on different surfaces or planes. To address this limitation, an enhanced method for object height estimation using a monocular image is proposed. The proposed approach enables height estimation for objects positioned on different but parallel planes. To achieve this, a ground point is first estimated, defined as the intersection of the target object and the reference object's plane. Using this ground point, the height of the target object is measured by calculating the ratio of the target object's height above the ground to the height of its plane above the ground. Furthermore, the improvement in estimation accuracy is mainly due to automated vanishing line detection based on gravity direction estimation. This reduces user dependency and stabilizes geometric estimation. Experimental results show that the proposed method consistently reduces estimation error compared to conventional approaches. In particular, the method demonstrated improved robustness, achieving substantially lower standard error across repeated trials, which indicates more stable and reliable performance under varying conditions.
Keywords:
computer vision
height estimation
height measurement
image metrology
monocular camera
vanishing point and line
Journal
J
IF:
1.8
Papers:
175
Citations:
0