| 논문명 | Know Your Limits: Accuracy of Long Range Stereoscopic Object Measurements in Practice |
|---|---|
| 저자(소속) | Peter Pinggera() |
| 학회/년도 | ECCV 2014, 논문 |
| 키워드 | |
| 데이터셋(센서)/모델 | |
| 관련연구 | |
| 참고 | Youtube |
| 코드 |
양안 비젼 응용 서비스 들은 장애물의 속도와 위치를 알기 위해서 높은 수준의 precision를 요구 한다. Modern applications of stereo vision, such as advanced driver assistance systems and autonomous vehicles, require highest precision when determining the location and velocity of potential obstacles.
Subpixel disparity accuracy in selected image regions is therefore essential.KITTI 같은 데이터 셋은 dense matching performance측정하는데 좋은 데이터셋이지만 local sub-pixel matching accuracy을 다루기에는 충분하지 않다.
Evaluation benchmarks for stereo correspondence algorithms, such as the popular Middlebury and KITTI frameworks, provide important reference values regarding dense matching performance, but do not sufficiently treat local sub-pixel matching accuracy.본 논문에서는 In this paper, we explore this important aspect in detail.
최신 스테레오 매칭 접근법들에 대하여 평가를 진행 하였다. We present a comprehensive statistical evaluation of selected state-of-the-art stereo matching approaches on an extensive dataset and establish reference values for the precision limits actually achievable in practice.
좋은 알고리즘 선택을 위한 가이드 라인 제공 We present guidelines on algorithmic choices derived from theory which turn out to be relevant to achieving this limit in practice.
Middlebury 데이터셋으로 시작으로 활발한 연구가 시작됨 Part of the practicability and performance of modern stereo vision algorithms can arguably be attributed to the seminal Middlebury benchmark study[27], which first provided a comprehensive framework for evaluation and enabled algorithm analysis and comparison.KITTI 데이터넷으로 더욱 발전함 Ten years later, the KITTI project [10] presented a new realistic and more challenging benchmark with stereo imagery of urban traffic scenes, triggering a new wave of improved stereo vision algorithms.위 데이터들은 dense stereo correspondence에 초점을 두고 있으며, dense & accurate GT 데이터를 필요로 한다. These major benchmark studies focus on dense stereo correspondence and are naturally required to provide both dense and accurate ground truth data.
성능 평가는 Algorithm performance is mainly judged by the percentage of pixels whose disparity estimates fall within a given accuracy threshold.