Intelligent Autonomous Inspections Using Deep Learning and Detection Markers
Abstract
Inspection of industrial and scientific facilities is a crucial task that must be performed regularly. These inspections tasks ensure that the facility’s structure is in safe operational conditions for humans. Furthermore, the safe operation of industrial machinery, is dependent on the conditions of the environment. For safety reasons, inspections for both structural integrity and equipment is often manually performed by operators or technicians. Naturally, this is often a tedious and laborious task. Additionally, buildings and structures frequently contain hard to reach or dangerous areas, which leads to the harm, injury or death of humans. Autonomous robotic systems offer an attractive solution to the automation of inspections. This is due to their ability to reach remote and dangerous areas. Consequently, significant research efforts have been made for the use Unmanned Aerial Vehicle (UAV) and drones as flight inspection units. Compared to Unmanned Ground Vehicle (UGV), drones offer significantly better acrobatic and agility performance. Needless to say, an autonomous drone must posses certain abilities to carry out inspections with little or no human intervention. Foremost, the drone must be able to position and orient itself towards inspection waypoints. Moreover, the drone must be able to use a map of the environment to identify and navigate to areas of interest on an inspection mission. These abilities alone do not provide the drone the autonomy and efficiency required to perform quality inspections within required time frames. The inspection problem is often framed as a way to maximize the amount of volume a drone can cover through the use of coverage path planning. Therefore, drone systems spend most of the time inspecting areas or objects that produce little to no value for identifying defects. In addition, more time and drone flights must be performed to cover an entire area. Thus, this thesis proposes a novel approach to perform intelligent inspections using deep learning and Quick Response (QR) codes to identify, prioritize and segment objects of interest in the context of an inspection. Additionally, a novel navigation architecture that can find and prioritize collision-free trajectories is proposed. This architecture proposes the use of hierarchical state machines to capture complex reactive behavior found on a task driven robot. In conjunction, the proposed systems solve the inspection problem by identifying, prioritizing, labelling and navigating towards areas or objects of interest.
Subject Area
Robotics|Artificial intelligence
Recommended Citation
Martinez Acosta, Alejandro, "Intelligent Autonomous Inspections Using Deep Learning and Detection Markers" (2022). ETD Collection for University of Texas, El Paso. AAI29997505.
https://scholarworks.utep.edu/dissertations/AAI29997505