(Dr. Kok-Meng Lee, advisor)
"Development and Analysis of an Absolute Three Degree of Freedom Vision Based Orientation Sensor"
The design of multi-degree-of-freedom actuators, such as ball joint-like spherical motors, is often hampered by the lack of accurate, non-contact orientation sensing methods suitable for use in a control system. Typically, specially designed mechanisms must be used to isolate and transfer specific motions to traditional single degree-of-freedom (DOF) sensors. A three DOF spherical motor, for example, requires a gimbal mechanism to isolate each axis of rotation for the use of incremental shaft encoders. This approach inevitably results in undesirable increases in mechanical complexity, friction and weight. Machine vision systems may overcome these difficulties with the possibility of highly accurate non-contact position sensing. This thesis presents the theoretical and practical requirements of a three DOF vision based orientation sensor, as well as the first investigation and demonstration of experimental hardware, software, and algorithms in an ongoing exploration of the sensor concept.
The sensor system consists of a digital camera and a specially structured
spherical grid. The grid is designed such that any position on its surface,
equivalent to its orientation with respect to a fixed position, is uniquely
determined by the line patterns at that position. The need for homing or
the potential for accumulated errors found in incremental encoders is overcome.
The algorithms necessary for defining the grid are presented, as well as
the design of a mechanism, which includes mechanical, electrical, and software
control components, used to produce grids. The orientation sensing itself
consists of a machine vision algorithm for deducing the orientation of
the grid by analyzing an image of ~e grid surface. Experimental results
and the effects of implementation requirements are presented.