Novel machine vision methods for outdoor and built environments

Abstract

Machine vision is potentially the most powerful sensing technique for intelligent automation. However, the existing techniques and algorithms only partially satisfy needs of complex environments (e.g. outdoor construction sites) mainly because of poor and/or unpredictably changing visibility conditions frequently encountered there (weather phenomena, mud and silt, partially buried objects, burst of intensive illumination, etc.). In this paper, we discuss principles of two methods that can significantly improve performances of vision system in such difficult conditions. The first method (based on gated imaging) is a specialized image acquisition technique while the second method addresses an important problem of object detection in complex environments. Both methods have been tested using images captured under visual conditions typically expected in such environments (e.g. many images are captured in high-turbidity water). The methods can be incorporated into more complex sensing systems, depending on needs and constraints of particular applications.