Surveillance System for Autonomous Survey Vehicles

Autonomous Survey Vessels, also known as Autonomous Survey Vehicles (ASV) are used for surveying of lakes in Norway. This Master thesis explains about development of a surveillance system for Autonomous Survey Vehicles. This system consists of three parts – detection of the objects in the direction of ASV, measurement of the obstacle distance and deviation of ASV away from the obstacle to avoid collision. The objects in the ASV path are classified and located in the image by using Raspberry Pi, PiCamera and machine learning with pre-trained object detection model from TensorFlow object detection repository. The object detection model is trained with custom data set to detect objects in lake environment. Distance of the obstacle in ASV path is estimated by measuring and combining data from Ultrasound and Laser sensor. The Ultrasound sensor measurement includes temperature and humidity correction. Influence of obstacle shape, surface and ambience light on sensor measurement is studied. By using information from object detection and obstacle distance measurement, ASV is steered away from obstacle to avoid collision.