Object-Centric Stereo Ranging for Autonomous Driving: From Dense Disparity to Census-Based Template Matching

Qihao Huang

cs.CVApr 9, 2026

Object-Centric Stereo Ranging for Autonomous Driving: From Dense Disparity to Census-Based Template Matching

Authors: Qihao Huang

Abstract

Accurate depth estimation is critical for autonomous driving perception systems, particularly for long range vehicle detection on highways. Traditional dense stereo matching methods such as Block Matching (BM) and Semi Global Matching (SGM) produce per pixel disparity maps but suffer from high computational cost, sensitivity to radiometric differences between stereo cameras, and poor accuracy at long range where disparity values are small. In this report, we present a comprehensive stereo ranging system that integrates three complementary depth estimation approaches: dense BM/SGM disparity, object centric Census based template matching, and monocular geometric priors, within a unified detection ranging tracking pipeline. Our key contribution is a novel object centric Census based template matching algorithm that performs GPU accelerated sparse stereo matching directly within detected bounding boxes, employing a far close divide and conquer strategy, forward backward verification, occlusion aware sampling, and robust multi block aggregation. We further describe an online calibration refinement framework that combines auto rectification offset search, radar stereo voting based disparity correction, and object level radar stereo association for continuous extrinsic drift compensation. The complete system achieves real time performance through asynchronous GPU pipeline design and delivers robust ranging across diverse driving conditions including nighttime, rain, and varying illumination.

Figures & tables

	Mono	Stereo	LiDAR	Radar
Depth type	Learned	Geometric	Active	Active
Cost ($)	10–30	30–80	450–8k	50–200
Range (m)	$<$ 80	$<$ 300+	$<$ 200	$<$ 250
Density	Dense	Dense	Sparse	Very sparse
Non-std obj.	Poor	Good	Good	Poor
Weather	Moderate	Good	Poor	Excellent

Table 1: Comparison of depth sensing modalities for autonomous driving. Stereo vision uniquely combines physics-based metric depth, low cost, and dense output.

Panel of Figure 2 — Figure 2: Fundamental paradigm comparison of three stereo matching methods. (a) Block Matching slides a fixed SAD window along epipolar lines for every pixel in the image. (b) Semi-Global Matching aggregates pixel-wise matching costs along 4–8 directions with smoothness penalties ( $P_{1},P_{2}$ ) for every pixel. Both produce dense disparity maps at $O(WH)$ cost. (c) Our Template Matching ( highlighted ) performs Census-based matching only within detected bounding boxes , processing 1–5% of image pixels while achieving higher accuracy at long range through evidence aggregation.

Panel of Figure 3 — Figure 3: Census Transform encoding and matching. Each pixel’s intensity is compared to the center pixel, producing a binary descriptor invariant to monotonic intensity changes. Matching cost is the Hamming distance (popcount of XOR) between left and right descriptors. Green cells: intensity $>$ center ( $\to 1$ ); red cells: $\leq$ center ( $\to 0$ ); orange: center pixel.

Panel of Figure 4 — Figure 4: Template Matching far/close dual-track strategy and forward-backward verification. Far objects (small bounding boxes) are matched as a single block at full resolution. Close objects (large bounding boxes) are subdivided into a grid, matched at scaled resolution, and robustly aggregated via sorted contiguous segment median. Both paths employ forward-backward verification (center inset): a match is accepted only if the backward search returns to near the original position.

Property	BM	SGM	Template
Output	Dense map	Dense map	Per-object
Matching cost	SAD	Census/MI	Census
Regularization	None	4-dir path	F-B verify
Radiometric inv.	Low	Medium	High
Long-range acc.	Low	Medium	High
Computation	High	High	Low

Table 2: Comparison of stereo matching methods.

Stereo Matching

Stereo matching estimates depth by finding corresponding points in paired images and measuring their disparity.

See all 57 papers

Horizon3D: Sparse Radar-Camera Fusion for Long-Range 3D Perception in Autonomous Driving
Jun 30, 2026Geonho Bang, Geunju Baek, Dongyoung Lee +2Camera-Radar Fusion Autonomous Driving
Seoul National University, Seoul, Republic of Korea
URS-Stereo: Uncertainty-Guided Residual Search for Real-Time Stereo Matching
Jul 7, 2026Pouya Sohrabipour, Chaitanya kumar reddy Pallerla, Dongyi WangStereo Matching
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

Jun 30, 2026J/K move · Enter open · S save

Horizon3D: Sparse Radar-Camera Fusion for Long-Range 3D Perception in Autonomous Driving

Geonho Bang, Geunju Baek, Dongyoung Lee, Wonjun Jeong+1

Seoul National University, Seoul, Republic of Korea

Camera-Radar Fusion Autonomous Driving 3D Object Detection Multimodal Sensor Fusion

Open paper AskPDF arXiv

Long-range 3D object detection is critical for safe autonomous driving at highway speeds, yet existing radar-camera fusion methods remain limited at extended ranges. BEV-based methods capture scene-level context but incur rapidly growing computation and often lose fine-grained object detail, while query-based methods are efficient but provide limited scene-level context. Temporal fusion further requires both multi-frame accumulation for sparse distant observations and object-level motion modeling for fast-moving objects. We propose Horizon3D, a sparse radar-camera fusion framework for long-range 3D object detection that combines Gaussian primitives with sparse BEV features. Horizon3D initializes Gaussian primitives at radar- and camera-estimated object keypoints using Keypoint-Guided Gaussian Initialization, refines them through Object-Centric Sparse Fusion, and splats them onto the BEV plane to fuse object-level detail with sparse radar BEV context. It further introduces Dual-Path Temporal Fusion, which aggregates temporal cues through a BEV path for scene-level accumulation and a Gaussian path for object-level motion propagation. Experiments on TruckScenes show that Horizon3D achieves state-of-the-art radar-camera 3D detection performance. On the validation set, it outperforms the previous best method by +3.0 NDS and +1.6 mAP while maintaining competitive inference speed.

Abstract

Stereo Matching

Explore similar work

Horizon3D: Sparse Radar-Camera Fusion for Long-Range 3D Perception in Autonomous Driving

URS-Stereo: Uncertainty-Guided Residual Search for Real-Time Stereo Matching

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