Various technologies were used to detect, track, and classify vessels on the Hudson River. Broadband radar was used to detect and track vessels. Visible light cameras, infrared cameras, and image processing techniques were used to detect, track, and classify vessels. Automatic Identification System (AIS) was used to track and classify vessels. The technologies, collectively referred to as the Integrated Technology System (ITS), were used in conjunction with each other to achieve synergies and to overcome individual system limitations. These limitations included a narrow field of view, false alarms, and misdetections. The suite of technologies successfully fulfilled its purpose. The radar was effective despite some errors. The cameras allowed for software development including automatic slewing and image processing. While AIS was considered the most reliable tool, it was determined not to be infallible. Future work includes integration of passive acoustics into the system and wake analysis for vessel detection.
This paper presents a new 3D scene reconstruction technique using the Unity 3D game engine. The method presented here allow us to reconstruct the shape of simple objects and more complex ones from multiple 2D images, including infrared and digital images from indoor scenes and only digital images from outdoor scenes and then add the reconstructed object to the simulated scene created in Unity 3D, these scenes are then validated with real world scenes. The method used different cameras settings and explores different properties in the reconstructions of the scenes including light, color, texture, shapes and different views. To achieve the highest possible resolution, it was necessary the extraction of partial textures from visible surfaces. To recover the 3D shapes and the depth of simple objects that can be represented by the geometric bodies, there geometric characteristics were used. To estimate the depth of more complex objects the triangulation method was used, for this the intrinsic and extrinsic parameters were calculated using geometric camera calibration. To implement the methods mentioned above the Matlab tool was used. The technique presented here also let’s us to simulate small simple videos, by reconstructing a sequence of multiple scenes of the video separated by small margins of time. To measure the quality of the reconstructed images and video scenes the Fast Low Band Model (FLBM) metric from the Video Quality Measurement (VQM) software was used. Low bandwidth perception based features include edges and motion.