Murat Haciomeroglu

Silhouette Extraction from Street View Images

This study addresses the issue of silhouette extraction of a street, and proposes two novel approaches to overcome this problem. The first, namely hybrid-stitching, considers the silhouette extraction as an image stitching problem and aims to use 2D street view images. The algorithm used in this method integrates a new composition technique into a conventional image stitching pipeline. The developed software using the proposed hybrid approach results in better stitching performances when compared with the popular stitching tools in the literature. Despite the results of the proposed method are better than the state-of-the-art image stitching techniques in many cases, they are not reliable enough to handle all of the street view image sets. Accordingly, a second solution has been proposed, including 3D location information, namely, 3D Silhouette Extraction Pipeline. The pipeline involves several techniques and post-processing steps to handle both the transformation and projection of the obtained point cloud, and the elimination of misleading location information. The results reveal that compared with the 2D solutions, the proposed algorithm is very effective and more reliable in silhouette extraction of a street, which is critical in urban transformation and environmental protection.

Fuzzy Logic Controlled Pedestrian Groups in Urban Environments

Populating an urban environment realistically with thousands of virtual humans is a challenging endeavour. Previous research into simulating the many facets of human behaviour has focused primarily on the control of an individual’s movements. However, a large proportion of pedestrians in an urban environment walk in groups and this should be reflected in a simulation. This paper, therefore, proposes three fuzzy logic engines in order to adjust the speed of the multi level groups in urban environments. The proposed nested multiple fuzzy logic engines maintain the balance between desired speed, main group and sub group configurations. Thus, a natural and non-rigid group-based locomotion is achieved in urban settings. The realism of the presented techniques is verified by comparing them with statistics acquired from a study on real human behaviour. It is shown that the inter-personal distances between group members, the speed of individuals and the speeds of pedestrians in groups are consistent with their real counterparts.

Distributing Pedestrians in a Virtual Environment

Distributing a collection of virtual humans throughout a large urban environment, where limited semantic information is available, poses a problem when attempting to create a visually realistic real time environment. Randomly positioning agents within an urban environment will not cover the environment with virtual humans in a plausible way. For example, areas of the urban space that are more frequently used should have a higher population density both at the start and during the simulation. It is infeasible to manually identify areas in the urban environment which should be crowded or sparsely populated when considering a scalable method, suitable for large environments. Consequently, this paper combines and extends techniques from spatial analysis and virtual agent behaviour simulations to develop a system capable of automatically distributing pedestrians in an urban environment. In particular, it extends the pointbased space syntax technique to enable the automatic analysis of a large urban environment in the presence of limited contextual information. This analysis specifies a set of population densities for areas in the environment and these values are used to initialise the locations of all the virtual humans in the environment. In addition to the initialisation stage the population densities in each area are consulted to ensure that the correct distribution of virtual humans is maintained throughout the simulation. The system is tested on an arbitrary section of a real city and comparisons of the characteristic parts of the test environment are correlated with the pedestrian movements.

Congestion-free multi-agent navigation based on velocity space by using cellular automata

In this study have we focused on two aspects of multi-agent simulations. The first is based on a finding in recent years, which is that a standalone global path does not always provide adequate multi-agent navigation in crowded scenarios. A global planner that is aware of other agent configurations and thus finds clearer paths is required for optimal navigation. In real life, usually, an agent is only aware of the area close to it. In this study, by taking into account this limitation, we propose a state-machine-based global planner that monitors agents’ close domains and, if required, modifies the path for congestion prevention. The second aspect is the coordination of local and global planners. Multi-agent navigation systems require both local steering and global path planning. As a matter of fact, these two systems should work in coordination since the output of one influences the other. A steering technique cannot always guarantee smooth and collision-free navigation while tracking the global planner’s path. Therefore, we propose a system that detects and rebuilds coordination between local and global path planners.

A path-based multi-agent navigation model

The quality of a crowd simulation model is determined by its agents’ local and global trajectory efficiency. While an agent-based model can accurately handle the local trajectories, global decisions usually are handled by a global path planner. However, most of the global path planning techniques do not consider other agents and their possible paths and the future global flow in the environment. In this paper, we propose a composite system that takes future agent configurations into account via a modified A* algorithm to create a global path plan and combines the global path plan with a local navigation model. We show that the agents using the proposed model intelligently plan their paths based on the dynamic configuration of the environment. In order to balance the performance vs. trajectory quality trade-off, we propose a hierarchical grid structure and discuss its effects on both trajectory quality and computational performance.

Real-time collision-free linear trajectory generation on GPU for crowd simulations

Crowd simulations are mostly employed to compose a background in the current scene. For such ambient crowds, it might be unnecessary to perform complex steering calculations. In this study, a steering-free crowd simulation which eliminates the computational cost arising from expensive steering maneuvers is presented. For this purpose, agents are assigned a linear trajectory that is guaranteed to be collision free before entering the simulation. These trajectories are calculated using readily available rendering pipeline of the GPU. To that end, existing agents’ bounding discs are rendered in a spatio-temporal manner as each one forms a straight 3D tube and a projection from a selected initial position is captured. Using the blank areas (holes) in this image, it is possible to determine a suitable constant velocity (a goal position and a speed). In experiments, we not only assess three different methods to choose one of the candidate solutions, but also compare our approach with an existing work. Test results reveal that our technique gives better results in both populating an empty environment with agents quicker and reaching a higher maximum number of agents than the existing method.

Hardware-accelerated dynamic clustering of virtual crowd members

In this study, a hardware‐accelerated dynamic clustering of moving virtual entities technique is proposed. By clustering virtual entities, both clustered and unclustered virtual agents became more aware of other agents’ topological configurations. Clustering is based on their continuously changing velocity and position vectors. The proposed clustering technique efficiently uses graphics processors parallel processing capabilities. Therefore, almost no additional central processing unit overhead is required to bring the clustering information into the simulation. In addition, in this paper, how cluster information can be used on top of the proposed virtual human steering technique is explained. The results show that by using the dynamic clustering, the number of collision in the simulation reduces, and the velocities of the the agents in the simulation increase. Copyright

Perceptual evaluation of maneuvering motion illusion for virtual pedestrians

Crowd simulations span a wide spectrum of application domains, most notably video games, evacuation scenarios, and the movie industry. However, it is not obligatory that all virtual populace applications have the primary objective of realistic simulation. In most instances, it is necessary and sufficient that viewers perceive the crowd as plausible. Even for a crowd consisting of agents navigating on linear trajectories without any maneuvers, visual motion illusion elicited by these trajectories might appear to be a natural consequence, causing them to be perceived as wriggling rather than straight. In this respect, we evaluate in this study whether simulated 3D human agents walking with constant, collision-free velocities, induce such a maneuvering motion illusion, aiming toward an efficient real-time crowd simulation. For this purpose, we recorded videos of virtual human crowds with different parameter combinations, such as the agent walking speed, crowd density, camera tilt angle, and camera distance. These videos were watched by human subjects who were instructed to mark the virtual agents who they thought had changed their gait directions. The analyzed results revealed that participants claimed the presence of maneuvering virtual agents in the videos, even though there were none in any of them. Spatial grouping of the markings highlighted that the participants mainly focused on the central area of the simulation environment, and spatiotemporal analysis of the click data also showed stronger evidence to such an illusion (see accompanying video). Furthermore, we found that all of the referred parameters have statistically significant main effects on the number of marked agents per watched video.

Combining GPU-generated linear trajectory segments to create collision-free paths for real-time ambient crowds

Abstract Ambient crowds are used widely in today’s computer games and movies. In this study, a real-time steering-free ambient crowd navigation technique is presented, which combines linear and constant-speed trajectory segments end-to-end to provide collision-free paths to simulated agents. This method seeks spatio-temporal space to find such trajectory segments by utilizing specialized features of the GPU. Each segment is required to come one after another, and at the transition points, the segments are restricted to have direction and/or speed changes within certain limits. Experiments revealed that the proposed multi-segment path technique can create denser crowds by simulating more agents (up to 53% on average and up to 44% as a maximum) than the single-segment path method existing in the literature. It is also shown in the experiments that the presented technique requires less time per navigated agent per time step (up to 98%) than a popular velocity-based microscopic method.

Automatic spatial analysis and pedestrian flow control for real-time crowd simulation in an urban environment

Distributing a collection of virtual humans throughout a large urban environment, where limited semantic information is available, poses a problem when attempting to create a visually realistic real-time environment. Randomly positioning agents within an urban environment will not cover the environment with virtual humans in a plausible way. For example, areas of the urban space that are more frequently used should have a higher population density both at the start and during the simulation. It is infeasible to manually identify all the areas in the urban environment which should be crowded or sparsely populated when considering a scalable method, suitable for large environments. Consequently, this paper combines and extends techniques from spatial analysis and virtual agent behaviour simulations to develop a system capable of automatically distributing pedestrians in an urban environment. In particular, it extends the Point-Based Space Syntax technique to enable the automatic analysis of a large urban environment in the presence of limited contextual information. This analysis specifies a set of population densities for areas in the environment and these values are used to initialise the locations of all the virtual humans in the environment. In addition to the initialisation stage the population densities in each area are consulted to ensure that the correct distribution of virtual humans is maintained throughout the simulation. The system is tested on an arbitrary section of a real city and comparisons of the characteristic parts of the test environment are correlated with the pedestrian movements.