Calibration-Free Vehicle Speed Estimation: A Monocular Keypoint-Template Approach

Gaofeng Su; Keya Li; Raja Sengupta; Kara M. Kockelman

cs.CVAug 17, 2026

Calibration-Free Vehicle Speed Estimation: A Monocular Keypoint-Template Approach

Authors: Gaofeng Su, Keya Li, Raja Sengupta, Kara M. Kockelman

Organizations: Department of Civil & Environmental Engineering The University of California, Berkeley · Department of Civil, Architectural and Environmental Engineering The University of Texas, Austin

Abstract

This paper proposes a calibration-free framework for reliably and effectively estimating vehicle speeds from monocular videos, without relying on roadway features, camera calibration, or roadway-feature-based reference objects. The proposed framework estimates vehicle speeds using a 36-keypoint vehicle template and a homography matrix updated at each frame. A YOLO-based keypoint detection module is trained on diverse datasets, and two estimation strategies are compared: keypoint-only tracking and warped optical flow with dense spatial aggregation. Speed is estimated by projecting displacements into metric space using the homography, with validation conducted on over 400 video clips from roadside and overhead datasets, covering speeds from 30 to 100 mph. The method achieves reliable speed estimation on the VS13 and BrnoCompSpeed datasets, with the warped optical flow method delivering MAEs of 15.0% and 9.7%, respectively, and 77.9% and 93.1% of estimates falling within +/-20% error. After applying a 10% trim to remove edge-of-frame outliers, performance improves to MAEs of 11.7% and 7.6%, with within-+/-20% accuracy increasing to 85.3% and 95.4%. This work addresses key limitations of existing vision-based approaches and enables low-cost and efficient speed enforcement using portable devices such as dashcams and smartphones, thereby supporting citizen-based enforcement programs for traffic safety.

Figures & tables

Figure 1: Piecewise planar approximation of a vehicle body. (Frame sampled and modified from the VS13 dataset Djukanović et al. (2022) ).

Figure 2: Illustration of speed estimation via warped optical flow for a tracked vehicle’s right facet. The vehicle graphic was generated using AI tools.

Figure 4: 36 semantic keypoints for the sedan are marked with blue circles (symmetric on the other side). Assuming a wheelbase of approximately 3 meters. Car graphics were adapted from Wikipedia. 842U (2010)

Panel of Figure 5 — Figure 5: Sample image from the VS13 dataset (left) and BrnoCompSpeed dataset (right), annotated using the speed estimation framework.

Data	Method	Group	N	Med.	MAE	RMSE	Thresholds
Data	Method	Group	N	Med.	MAE	RMSE	$\pm$ 5%	$\pm$ 10%	$\pm$ 20%	$>$ 50%
VS13	KP	All	4,317	-0.9%	20.8%	178.4%	46.7%	72.5%	87.6%	4.4%
		Sedan	2,055	-0.7%	24.3%	243.0%	46.0%	73.3%	89.7%	3.2%
		NoSed	2,262	-1.0%	17.7%	84.2%	47.3%	71.7%	85.7%	5.6%
	OF	All	4,578	-6.0%	15.0%	53.5%	37.3%	58.8%	77.9%	5.3%
		Sedan	2,195	-6.1%	14.3%	22.1%	35.4%	56.8%	76.1%	4.8%

Table 1: Summary of speed estimation accuracy for VS13 and BrnoCompSpeed datasets.

Panel of Figure 6 — Figure 6: Error Percentage vs. Frame Index for VS13 and BrnoCompSpeed datasets

Data	Method	Group	N	Med.	MAE	RMSE	Thresholds
Data	Method	Group	N	Med.	MAE	RMSE	$\pm$ 5%	$\pm$ 10%	$\pm$ 20%	$>$ 50%
VS13	KP	All	3,334	-1.2%	14.3%	170.6%	54.0%	81.0%	94.0%	2.1%
		Sedan	1,584	-1.2%	18.5%	240.9%	52.0%	79.9%	94.1%	1.8%
		NoSed	1,750	-1.3%	10.5%	54.1%	55.7%	82.0%	93.9%	2.3%
	OF	All	3,555	-5.1%	11.7%	57.4%	43.0%	66.2%	85.3%	2.3%
		Sedan	1,692	-5.4%	11.6%	17.7%	40.1%	62.4%	81.7%	2.4%

Table 2: Summary of speed estimation accuracy for VS13 and BrnoCompSpeed datasets with 10% trimming.

Panel of Figure 7 — Figure 7: Error Percentage vs. Speed for VS13 and BrnoCompSpeed datasets

Physics-Grounded Monocular Vehicle Distance Estimation Using Standardized License Plate Typography
Apr 14, 2026Manognya Lokesh Reddy, Zheng LiuCollision Avoidance Metric Depth Estimation
Department of Computer and Information Science University of Michigan-Dearborn Dearborn, Michigan, USA · Department of Industrial and Manufacturing Systems Engineering University of Michigan-Dearborn Dearborn, Michigan, USA
Infrastructure-based Monocular 3D Vehicle Localization Framework with Experimental Validation
Sep 1, 2026Akos T. Kopeczi-Bocz, Tian Mi, Gabor Orosz +1Autonomous Driving Perception Monocular 3D Object Detection
Department of Applied Mechanics, Budapest University of Technology and Economics, Budapest, H-1111, Hungary · Department of Mechanical Engineering, University of Michigan, Ann Arbor, MI 48109, USA · Department of Mechanical Engineering, the Department of Civil and Environmental Engineering, and the Transportation Research Institute at the University of Michigan, Ann Arbor, MI 48109, USA
Smartphone-Based Method for Automated Speed Enforcement
Sep 24, 2026Keya Li, Jahnavi Malagavalli, Lamha Goel +2Intelligent Transportation Systems Automatic License Plate Recognition
Department of Civil, Architectural and Environmental Engineering The University of Texas at Austin · Department of Computer Science The University of Texas at Austin · Chandra Department of Electrical and Computer Engineering The University of Texas at Austin

Apr 14, 2026J/K move · Enter open · S save

Physics-Grounded Monocular Vehicle Distance Estimation Using Standardized License Plate Typography

Manognya Lokesh Reddy, Zheng Liu

Department of Computer and Information Science University of Michigan-Dearborn Dearborn, Michigan, USA · Department of Industrial and Manufacturing Systems Engineering University of Michigan-Dearborn Dearborn, Michigan, USA

Collision Avoidance Metric Depth Estimation Monocular Depth Estimation Autonomous Driving Perception

Open paper AskPDF arXiv

Accurate inter-vehicle distance estimation is a cornerstone of Advanced Driver Assistance Systems (ADAS) and autonomous driving. While LiDAR and radar provide high precision, their high cost prohibits widespread adoption in mass-market vehicles. Monocular camera-based estimation offers a low-cost alternative but suffers from fundamental scale ambiguity. Recent deep learning methods for monocular depth achieve impressive results yet require expensive supervised training, suffer from domain shift, and produce predictions that are difficult to certify for safety-critical deployment. This paper presents a framework that exploits the standardized typography of United States license plates as passive fiducial markers for metric ranging, resolving scale ambiguity through explicit geometric priors without any training data or active illumination. First, a four-method parallel plate detector achieves robust plate reading across the full automotive lighting range. Second, a three-stage state identification engine fusing optical character recognition text matching, multi-design color scoring, and a lightweight neural network classifier provides robust identification across all ambient conditions. Third, hybrid depth fusion with inverse-variance weighting and online scale alignment, combined with a one-dimensional constant-velocity Kalman filter, delivers smoothed distance, relative velocity, and time-to-collision for collision warning. Baseline validation on a controlled static dataset reproduces a 2.3% coefficient of variation in character height measurements and a 36% reduction in distance-estimate variance compared with plate-width methods from prior work.

Abstract

Explore similar work

Physics-Grounded Monocular Vehicle Distance Estimation Using Standardized License Plate Typography

Infrastructure-based Monocular 3D Vehicle Localization Framework with Experimental Validation

Smartphone-Based Method for Automated Speed Enforcement