Study on a Location Method for Bio-objects in Virtual Environment Based on Neural Network and Fuzzy Reasoning

Abstract

In picking manipulator location system, it is the actual challenge that the position of a bio-object is exactly determined in complex and uncertain environment. An accurate location method for a bio-object in virtual environment based on binocular stereo vision was studied. Firstly, an experiment platform for a bio-object and picking manipulator location based on binocular stereo vision was established, and the imaging of a bio-object and environment were gained by binocular stereo vision. Secondly, the causes of errors existing in the process of handling the bio-object and environment imaging were analyzed. An approach was point out that the vision errors were amended by means of building a neural network and a corresponding training system. Thirdly, considering there are many complex and uncertain factors existing in natural environment, fuzzy rule sets for ascertain the bio-object position were extracted and modeled. A fuzzy reasoning model which has multi-rules and multi-inputs is created based on experience of agricultural experts, which makes the location system more accurate and intelligent. Finally, a location simulation system was developed by VC++ and EON SDK, which can simulate a process of picking manipulator automatically locating to a bio-object.