INDEX 31th ICTCT Conference On the track of future urban mobility: safety, human factors and technology 1. INTRODUCTION 2. OBJECTIVE DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING 3. METHODOLOGY ENVIRONMENTS BASED ON POINT CLOUDS 3.1. Visibility analysis L. Díaz-Vilariño , C. Silva, P. Arias 3.2. 2D Isovist Field 3.3. 3D Isovist Field 4. EXPERIMENTS & RESULTS 5. CONCLUSIONS Porto, October 26th, 2018 DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS 1 INDEX 1. INTRODUCTION What is a point cloud ? 1. INTRODUCTION 2. OBJECTIVE 3. METHODOLOGY 3.1. Visibility analysis 3.2. 2D Isovist Field 3.3. 3D Isovist Field 4. EXPERIMENTS & RESULTS 5. CONCLUSIONS DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS 2 DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS 3
1. INTRODUCTION 1. INTRODUCTION What is a point cloud ? What is a point cloud ? Principle of measurement 8.988, 3.345, 1.235 Total Station vs Laser Scanner (x, y, z) Geometry • Intensity • Colour • (…) • Measurement ‘point by point’ Measurement rate: 1,000,000 points/sec DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS 4 5 1. INTRODUCTION 1. INTRODUCTION What is a point cloud ? What is a point cloud ? Point clouds -Urban and interurban infrastructures Platforms � Represent existing reality Mobile Mobile low-weight Mobile Static • • • • � High quality Accuracy • Resolution • � Huge amount of (big) data Methodologies to automatically process data and extract useful information DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS 6 DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS 7
1. INTRODUCTION 1. INTRODUCTION What is a point cloud ? What is a point cloud ? Point clouds -Urban and interurban infrastructures Point clouds -Urban and interurban infrastructures � Represent existing reality � Represent existing reality � High quality � High quality Autonomous vehicles Accuracy Accuracy • • Resolution Resolution • • Low-quality data • Perceiving the as-built environment � Huge amount of (big) data • � Huge amount of (big) data Real-time applications (obstacle detection, • etc.) Terrestrial Laser Scanners Methodologies to automatically Methodologies to automatically High-quality data • process data and extract useful process data and extract useful • Analysing and understanding the as- information information built environment DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS 8 9 INDEX 2. OBJECTIVE 1. INTRODUCTION To develop a methodology to the automatic analysis of visibility in urban 2. OBJECTIVE 3. METHODOLOGY crossings based on point clouds 3.1. Visibility analysis 3.2. 2D Isovist Field 3.3. 3D Isovist Field 4. EXPERIMENTS & RESULTS 5. CONCLUSIONS DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS 10 DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS 11
2. OBJECTIVE INDEX 1. INTRODUCTION To develop a methodology to the automatic analysis of visibility in urban 2. OBJECTIVE crossings based on point clouds 3. METHODOLOGY Why ? 3.1. Visibility analysis 3.2. 2D Isovist Field 3.3. 3D Isovist Field Poor visibility is highlighted as an important cause of road accidents • 4. EXPERIMENTS & RESULTS 3D realistic models of the as-built environment are essential for accurate analysis • 5. CONCLUSIONS DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS 12 13 3. METHODOLOGY 3. METHODOLOGY GENERAL WORKFLOW 3.1. VISIBILITY ANALYSIS Visibility analysis (2D or 3D) is performed for 3.1.1. Ray-tracing algorithm 2D Isovist Ray-tracing Occlusion each position of an observer , considering: Scene is discretized in LoSs covering the field of view algorithm detection Visual angle • Point cloud Maximum line of sight • Eye gaze direction • 3D Isovist Ray-tracing algorithm to detect occlusions (obstacles interfering with LoSs) DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS 14 DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS 15
3. METHODOLOGY 3. METHODOLOGY 3.1. VISIBILITY ANALYSIS 3.2. 2D ISOVIST FIELD Isovist theory (1979) defines an isovist as a polygon representing the space seen form one • Visibility analysis (2D or 3D) is performed for 3.1.2. Occlusion condition observation point each position of an observer , considering: Existence of points intersecting with LoSs (within a Visual angle • buffer of radius r) Space view defined by: • • Isovist defined by occlusion points in 2D Maximum line of sight • visual angle (2D) Eye gaze direction • • maximum line of sight • eye gaze direction • Ray-tracing algorithm to detect occlusions Figure. Occlusion condition (obstacles interfering with LoSs) Figure. Isovist Figure. 2D human visual field. DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS 16 17 3. METHODOLOGY INDEX 3.3. 3D ISOVIST FIELD 1. INTRODUCTION Extending the Isovist theory from 2D to 3D based on the direct use of point clouds. • 2. OBJECTIVE Space view defined by: • • Isovist defined by occlusion points in 3D 3. METHODOLOGY visual angle (3D) • maximum line of sight • eye gaze direction 3.1. Visibility analysis • 3.2. 2D Isovist Field 3.3. 3D Isovist Field 4. EXPERIMENTS & RESULTS 5. CONCLUSIONS Figure. 3D human visual field. DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS 18 DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS 19
4. EXPERIMENTS & RESULTS 4. EXPERIMENTS & RESULTS Case study 1 – Largo do Padrão (Porto) Case study 1 – Largo do Padrão (Porto) 6 hours • 15,4 kms • 33.000 10 6 points • Figure: Largo do Padrão (Google Maps) Figure: Point Cloud visualized in false colour by height (534203 points) DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS 20 21 4. EXPERIMENTS & RESULTS 4. EXPERIMENTS & RESULTS 2D ISOVISTS 2D ISOVISTS Parameters Value Observer height 1.5 m Max. Line of Sight 20 m Visual angle -30º to 30º Area of visibility: 45.99 m 2 Area of visibility: 61.69 m 2 Area of visibility: 56.67 m 2 Horizontal Ang. Res. 5º A B C DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS 22 DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS 23
4. EXPERIMENTS & RESULTS 4. EXPERIMENTS & RESULTS 3D ISOVISTS 3D ISOVISTS Parameters Value Consecutive 2D Isovists Observer height 1.5 m Max. Line of Sight 20 m 3D Isovist (boundary of stop points) Visual angle -30º to 30º Horizontal/Vertical 5º Ang. Res. DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS 24 25 INDEX 5. CONCLUSIONS 1. INTRODUCTION • A methodology to automatically generate 2D and 3D isovists from point clouds in urban scenes. 2. OBJECTIVE 3. METHODOLOGY • Visibility is studied for the space view defined by visual angle , maximum line of sight and eye gaze direction . 3.1. Visibility analysis 3.2. 2D Isovist Field • Results show how obstacles interfere in the visibility from a 3D perspective (understand the perception and behaviour in urban scenes) 3.3. 3D Isovist Field 4. EXPERIMENTS & RESULTS • Results depend on data quality , specially in terms of completeness and noise . 5. CONCLUSIONS DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS 26 DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS 27
31th ICTCT Conference On the track of future urban mobility: safety, human factors and technology DRIVER VISIBILITY ASSESMENT ON URBAN CROSSING ENVIRONMENTS BASED ON POINT CLOUDS L. Díaz-Vilariño , C. Silva, P. Arias lucia@uvigo.es Porto, October 26th, 2018
Recommend
More recommend
