单目相机空间定位文献阅读

关键词

• 单目相机：Monocular camera
• 单目相机空间定位：Monocular camera geolocation
• 大范围单目PTZ相机空间定位：Large-range Monocular PTZ-camera geolocation
• 大范围(Large-range)、全方位(omnidirectional)、低配准点（less world-image pairs）、高适应性(High adaptability：无需内外方位元素标定、对硬件安装质量要求低)、误差改进(Error distribution improvement)
• 配准点：world-image pairs, image to world point correspondences
• 枪机：central perspective projection camera
• PTZ相机： pan-tilt-zoom (PTZ) camera
• 全景相机：omnidirectional camera
• 监控相机视频：surveillance camera video
• 三维目标检测：3D Object detection
• 视频增强GIS：GIS augmented video surveillance
• 单应矩阵：homography matrix (The matrix representing the imaging transformation is estimated from image-to-world point correspondences. )¹
• 以视频查地图（projection）：image-to-world point correspondences
• 以地图查视频（back-projection）：world-to-image point correspondences
• 观测目标：target
• 相机视中心线姿态角度：指当前ptz相机的姿态角度，此时也是视中心线在ptz相机坐标系上的相对值，该角度值各取值范围；
• 观测目标视中心改正后姿态角度：经角度改正后，观测目标位于视中心线上时的ptz相机姿态角度

1999-IVC 《A plane measuring device》 300引用率 2022/2/4阅读 晴 大年初四 天气很好

0.Note

1.Cite

Criminisi, A. (2000). “Single-view metrology.” International Journal of Computer Vision.

2.GAP & solution

3.概念

3DGIS的优势

4.经句

• 主旨句：

5.总结

2000-书 《Multiple view geometry in computer vision》 30000万引用率 SLAM大神书

0.Note

Hartley, R. and A. Zisserman (2000). Multiple view geometry in computer vision, Cambridge university press.

1.Cite

2.GAP & solution

3.概念

3DGIS的优势

4.经句

• 主旨句：

5.总结

2000-IJCV 《Single-view metrology》1000引用率 2021/2/3阅读 阴 大年初三 感冒才好

0.Note

1.Cite

Criminisi, A. (2000). “Single-view metrology.” International Journal of Computer Vision.

2.GAP & solution

3.概念

3DGIS的优势

4.经句

• 主旨句：

5.总结

2010-IJGIS 《GIS augmented video surveillance》

0.Note

40引用率 视频GIS鼻祖文章，主要讲述如何融合地理要素和视频

阅读时间： 2021/1/24晨 雨 春节临近

1.Cite

Milosavljević, A., et al. (2010). “GIS-augmented video surveillance.” International Journal of Geographical Information Science 24(9): 1415-1433.

2.GAP & solution

随着监控路数的提升，传统多屏组合监控的方式的可操作性逐步降低，尤其是对安保人员提出了更高、更复杂的技能要求，需要从认知心理的角度识别视频画面中对象的空间位置、活动方向等。这显然是难以实现的。为了可以实现多屏的空间定位、跟踪及活动分析，SD提出了GIS是视频监控的通用参考框架，其有利于提供更多的空间语义信息。我们解决该问题所依赖途径是使用***方法。该方法有两点好处###。
A typical system of conventional video monitoring connects each video camera directly to a corresponding display screen. Therefore, we have as many screens as video cameras. In these kinds of systems, serious problems can occur when the scale of the monitoring system grows larger than human capacity. Security personnel (安保人员) must mentally map each surveillance monitor image to the corresponding place in the real world, and this complicated skill (复杂技能) requires experience and training (KAWASAKI, N. and TAKAI, Y., 2002. Video Monitoring System for Security Surveillance based on Augmented Reality, In Proceedings of the 12th International Conference on Artificial Reality and Telexistence, 4-6 December 2002, Tokyo, Japan, 180-181). To enable multi-camera coordination and tracking, Sankaranarayanan and Davis (SANKARANARAYANAN, K. and DAVIS, J.W., 2008. A Fast Linear Registration Framework for Multi-Camera GIS Coordination, In Proceedings of the 5th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS’08), 1-3 September 2008, Santa Fe, NM, 245-251.) emphasised the importance of establishing a common reference frame to which each of these cameras can be mapped. They suggested the use of GIS as a common frame of reference because it not only provides a solid ground truth, but more importantly provides spatial semantic information (e.g., locations of roads, buildings, sensitive areas, etc.) for use in applications such as tracking and activity analysis.
（另起一段）
Our solution to this problem relies on use of augmented reality techniques applied to GIS. In this approach, a GIS stores the necessary geospatial and contextual information about features that can be identified from a camera image. Furthermore, a GIS-based approach enables the inverse task of selecting and pointing an appropriate camera to some georeferenced feature or event.

3.概念

概念：3DGIS的优势
The advantages of 3D GIS over 2D GIS raise from the fact that it enables
visualisation and understanding of terrestrial phenomena and features that are only discernible in three dimensions. It also makes better presentations to those with little or no experience within the mapping.
概念：虚拟现实增强
Registration in augmented reality is a process that merges virtual objects generated by a computer with real-world images captured by a camera.
概念：PTZ相机
PTZ is an abbreviation for Pan-Tilt-Zoom, and in the terminology of video surveillance, it indicates cameras that can rotate in the
horizontal (pan) and vertical planes (tilt) and change their level of magnification (zoom).
PTZ相机安装位置需要做测量
When a PTZ camera is in the role of observer, the first group of parameters is fixed and determined by the camera mounting position. These parameters need to be measured (using GPS for example) and provided to the system for further use.
方法：注册的途径
In this paper, we present a method for registration of geospatial data applicable to outdoor video surveillance systems consisting of several PTZ cameras. Registration is based on transforming these relative camera view parameters into the absolute position, orientation, and field of view required by the 3D GIS. Once the 3D GIS and camera views are aligned, it is possible to identify geospatial objects from the camera image 视频查地图要素, as well as to overlap the virtual scene with the real one. Inverse transformation of the view parameters allows for selecting and pointing the appropriate camera by some georeferenced feature or event 地图要素查视频.
应用和价值
We suggested application of such system in emergency situation management and urban planning.

4.经句

• 主旨句：Research presented in this paper deals with the integration of 3D GIS and video surveillance systems.
• 主旨句：In this paper, we have presented the GeoScopeAVS system that integrates GIS and video surveillance for real-time retrieval of information about viewed geospatial objects.
• 主旨句：By some characteristics, GeoScopeAVS can be categorised in a group of monitorbased outdoor augmented reality systems.
• 主旨句：Using the 3D subobject paradigm, 3D and Camera views are
  enriched with the additional data representation model.

其他
本文也给出了如何写一个系统的基本形式，GeoScopeAVS。
本文给出了视觉变换的计算过程。

5.总结

基于相机内参标定和相机透视变换的方式依赖于准确的相机地理坐标位置和姿态信息。这些信息的准确性首先受限于相机安装时的规范性（如：不能出现水平失衡），其次，需要通过精准的户外测量。这种方式在理想条件下是容易实现的，如：相机的各类参数已知、相机的数量少、相机安装条件便于测量。然而，随着相机数量的增加、相机安装高度提升和安装位置复杂，准确的相机准确的地理坐标位置（GPS位置）和姿态信息获取面临着巨大的挑战，具体表现为：①很多相机安装高度、位置不便于测量，如：高空瞭望相机、封闭的室内相机；②由于高空相机安装规范性不足，很多相机安装并不达标。这在很大程度上也限制了三维视频GIS的实用化和应用。因此，研究不依赖准确的相机地理坐标位置和姿态信息的三维视频GIS方法对于推动视频增强GIS表达具有重要的意义。

2011-南师大博士论文《单幅图像几何量测方法》

0.Note

1.Cite

王美珍 (2011). 单幅图像中地物目标几何量测研究, 南京师范大学.

GAP & solution

当前图像获取的方式呈现多元化趋势，并且也无法得到摄影设备参数。不同来源的图像间摄影基线，无法事先控制，这都加大了双目视觉应用中图像匹配、相机自标定等过程难度。因此，破除传统摄影测量来自“双眼视觉”的束缚，发展基于单幅图像的量测技术，可有效利用单幅图像中蕴含的几何特征,避免了立体视觉中图像之间匹配、相机自标定等经典难题，将成为图像量测发展的重要趋势。

概念

单幅图像测量的起源研究：Criminisi(1999)和Hartley(2000)等系统地对单幅图像几何量测、单视图几何基础理论、方法作了总结和分析，主要包括平面与图像之间的映射、三维空间与图像之间的映射、图像变换的层次及层次变换保持的不变量等,为后期的研究奠定了基础。

单目视觉研究：单目视觉研究可广泛应用于：单幅图像几何量测、单幅图像相机标定、图像三维重建。三者的共同特点是都以图像中的几何信息作为线索,以图像与现实空间之间的成像关系为纽带,三者不同在于,单幅图像几何量测旨在获得图像中对象的几何尺寸,单幅图像相机标定旨在获得拍摄图像的相机参数,是三维重建的核心步骤,而三维重建则主要恢复图像的度量性质。

三维单应矩阵：三维相机矩阵有11个自由度,因此需要11个方程,由于每组
对应点可以确定两个方程,为了求解此单应矩阵,至少需要六组不退化的对应的图像点和空间点。当对应点的对数大于6对时,可用求其超定解²。

经句

• 主旨句：

总结

2014-ITITS Vision-only localization

0.Note

1.Cite

Lategahn, H. and C. Stiller (2014). “Vision-only localization.” IEEE Transactions on intelligent transportation systems 15(3): 1246-1257.

2.GAP & solution

3.概念

3DGIS的优势

4.经句

• 主旨句：

5.总结

2016-IJGIS 《Integration of GIS and video surveillance》

0.Note

1.Cite

Milosavljevic, A., et al. (2016). “Integration of GIS and video surveillance.” International Journal of Geographical Information Science 30(9-10): 2089-2107.

2.GAP & solution

3.概念

3DGIS的优势

4.经句

• 主旨句：

5.总结

2016-MVA 《Continuous localization and mapping of a ptz camera for wide area tracking》

0.Note

1.Cite

Lisanti, G., et al. (2016). “Continuous localization and mapping of a pan—tilt—zoom camera for wide area tracking.” Machine Vision and Applications.

2.GAP & solution

3.概念

3DGIS的优势

4.经句

• 主旨句：

5.总结

Lisanti, G., et al. (2016). “Continuous localization and mapping of a pan—tilt—zoom camera for wide area tracking.” Machine Vision and Applications.

Drawback:
these solutions are domain-specific and have no general applicability.
fiducial markers are likely to be occluded and impair the quality of tracking.

The main drawback of all these methods is that they assume that the scene is almost stationary and changes are only due to camera motion, which is a condition that is unlikely to happen in real contexts.

Beyond the fact that these solutions are domain-specific and have no general applicability, the main drawback is that fiducial markers are likely to be occluded and impair the quality of tracking.
The main contributions of the solution proposed are:
– We define a method for on-line PTZ camera calibration that jointly estimates the pose of the camera, the focal length and the scene landmark locations. Under reasonable assumptions, such estimation is Bayes-optimal, is very robust to zoom and camera motion and scales beyond thousands of scene landmarks. The method does not assume any temporal coherence between frames but only considers the information in the current frame.
– We provide an adaptive representation of the scene under observation that makes PTZ camera operations independent of the changes of the scene.
– From the optimally estimated camera pose we infer the expected scale of a target at any image location and compute the relationship between the target position in the 2D image and the 3D world plane at each time instant.
Differently from the other solutions published in the literature like [4], [7], [8] and [9], our approach allows performing on-line PTZ camera calibration also in dynamic scenes. Estimation of the relationship between positions in the 2D image and the 3D world plane permits more effective target detection, data association and real-time tracking. Some of the ideas for calibration contained in this paper were presented with preliminary(初步) results under simplified assumptions in [20,21]. Targets were detected manually in the first frame of the sequence and the scene was assumed almost static through time. Therefore we could not maintain camera calibration over hours of activity, neither support rapid camera motion.

2017-IJGI 《A method for estimating surveillance video georeferences》

0.Note

1.Cite

Milosavljević, A., et al. (2017). “A method for estimating surveillance video georeferences.” ISPRS International Journal of Geo-Information 6(7): 211.

2.GAP & solution

3.概念

3DGIS的优势

4.经句

• 主旨句：

5.总结

2020-ERE 《A monocular wide-field vision system for geolocation with uncertainties in urban scenes》3引用率 2021/2/3阅读 阴 大年初三 感冒才好

1.Cite

Arroyo, S. I., et al. (2020). “A monocular wide-field vision system for geolocation with uncertainties in urban scenes.” Engineering Research Express 2(2): 025041.

2.GAP & solution

3.概念

3DGIS的优势

4.经句

• 主旨句：

5.总结

2021-ITIP 《MGG: Monocular Global Geolocation for Outdoor Long-Range Targets》

0.Note

1.Cite

Gao, F., et al. (2021). “MGG: Monocular Global Geolocation for Outdoor Long-Range Targets.” IEEE Transactions on Image Processing 30: 6349-6363.

2.GAP & solution

3.概念

3DGIS的优势

4.经句

• 主旨句：

5.总结

---

1. Criminisi, A., et al. (1999). “A plane measuring device.” Image & Vision Computing 17(8): 625-634. ↩︎

2. Hartley, R. and A. Zisserman (2000). Multiple view geometry in computer vision, Cambridge university press. ↩︎