Placing Visual Sensors Using Heuristic Algorithms for Bridge Surveillance

Abstract

This study addresses the camera placement problem for bridge surveillance and proposes solutions that minimize the cost while satisfying the minimum coverage level. We discuss the field of view of cameras in the three-dimensional space. We also consider occlusions, the characteristics of surveillance targets, and different pan-tilt-zoom cameras in the visibility test. To solve the camera placement problem while minimizing the total cost, we propose a genetic algorithm (GA) and a uniqueness score with a local search algorithm (ULA). Problem sets for a large-scale dimension scenario are generated based on the data of actual bridges in the Republic of Korea. For three simulation sets and a case study of Samoonjin Bridge, the proposed ULA yields better results than GA.

Optimizing Fault-Tolerant Quality-Guaranteed Sensor Deployments for UAV Localization in Critical Areas via Computational Geometry

“…Or that a deployment which covers most of the volume of the RoI still leaves drones the sufficient amount of freedom to move undetected via, e.g., narrow, worm-shaped corridors (which, e.g., could form only after the sabotage of a single sensor), and approach, still undetected, a critical target, thus causing serious damage. Note that, being able to determine coverage point by point, as in [3], [4], and [5] in a discrete space is not enough to study this kind of properties of a deployment.…”

“…Several approaches consider the problem of monitoring a 3-D space, but with significant limitations affecting the applicability of these approaches to localizing small UAVs in large critical areas, such as those considered here. In [5], [11], [14], [27], [28], and [29], the candidate points for placement all lie on a 3-D surface; in [3], [4], [30], [31], [32], and [33], admissible sensor positions or points to be monitored (or both) belong to finite sets in the 3-D space. In particular, in most existing 3-D approaches, the RoI is discretized in cells.…”

“…For instance, mounting a heavy sensor on a wall costs more than mounting it on a roof, which in turn costs more than placing it on the ground. References [3], [5], [27], and [35] assume sensors have a fixed cost, while [14] and [36] assign to each sensor a cost based only on the altitude and the roughness of the terrain in its position. Conversely, our approach handles this issue as a first-class citizen.…”

“…1) Leonardo Da Vinci International Airport: This is a prototypical example of a critical area, with a total surface of approximately 16 km 2 , most of which is taken by the three 4-km long runways. We created a simplified 3-D model of the RoI as the union of 11 polyhedra with a height of 100 m. The total volume to be monitored is thus 1.6 km 3 . The case study presents many obstacles, modeled with 52 polyhedra, ranging from large buildings, such as hangars and terminals to small service buildings.…”

Sensor Placement Optimization of Visual Sensor Networks for Target Tracking Based on Multi-Objective Constraints

“…The VSNs are distributed perception networks that can track targets and achieve wide-area coverage, and they are composed of multiple smart camera nodes. At the moment, VSNs are widely used in surveillance [3,4], object detection [5], indoor patient monitoring [6], autonomous driving [7], smart city [8] and Unmanned Aerial Vehicles [9,10]. But the emergence of VSNs also brings with it a number of difficulties.…”