If humans represent such knowledge the rule for ball

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: e considered a ball? If humans represent such knowledge, the rule for ball circularity may look like If circularity is HIGH then object is a ball with HIGH confidence High circularity is a preferred property of balls. Such knowledge representation is very close to common sense representation of knowledge, with no need for exact specification of the circularity/non-circularity threshold. Fuzzy rules are of the form If X is A then Y is B Where X and Y represent some properties and A and B are linguistic variables. Semantic nets 4 Faculty of Engineering Robotics Technology MECH 4041 B. Eng (Hons.) Mechatronics S. Venkannah Mechanical and Production Engineering Department Semantic nets are a special variation of relational data structures. The semantics distinguish them from general nets—semantic nets consist of objects, their description, and a description of relations between objects. Logical forms of knowledge can be included in semantic nets, and predicate logic can be used to represent and/or evaluate the local information and local knowledge. Semantic nets can also represent common sense knowledge that is often imprecise and needs to be treated in a probalistic way. Semantic nets have a hierarchical structure; complex representations consist of less complex representations, which can in turn be divided into simpler ones, etc. relations between partial representations are described at all appropriate levels. Evaluated graphs are used as a semantic net data structure; nodes represent objects and arcs represent relations between objects. The following definition of a human face is an example of a simple semantic net; a face is circular part of the human body that consists of two eyes, one nose, and one mouth one eye is positioned left of the other eye the nose is between and below the eyes the mouth is below the nose an eye is approximately circular the nose is vertically elongated the mouth is horizontally elongated. The semantic net representing this knowledge is shown in FIG 7.3 Frames, scripts Frames provide a very general method for knowledge representation which may contain all the knowledge representation discussed above. They are sometimes called scripts because of their similarity to film scripts. Frames are suitable for representing common sense knowledge under specific circumstances. Consider a frame called plane_start; this frame may consist of the following sequence of actions: start the engines 5 Faculty of Engineering Robotics Technology MECH 4041 B. Eng (Hons.) Mechatronics S. Venkannah Mechanical and Production Engineering Department taxi to runway increase RPMs of engines to maximum travel along runway increasing speed fly Assuming this frame represents knowledge of how planes usually start, the situation of a plane standing on a runway with engines running causes the prediction that the plane will start in a short time. The frame can be used as a substitute for missing information which may be extremely important in vision related problems. Assuming that one part of the runway is not visible form the observation point, using the plane_start frame, a computer vision system can overcome the lack of continuous information between the plane moving at the beginning of the runway and flying when it next appears. If it is a passenger plane, the frame may have additional items such as time of departure, time of arrival, departure city, arrival city, airline, flight number, etc.. because in a majority of cases it makes sense to be interested in this information if we identify a passenger plane. From the formal point of view, a frame is represented by a general semantic net accompanied by a list of relevant variables, concepts, and concatenation of situations. No standard form of frames exists. Frames represent a tool for organizing knowledge in prototypical objects, and for description of mutual influences of objects using stereotypes of behavior in specific situations. Geometric properties: Given the binary image, it is possible to compute certain geometric properties (and in the case of multiple component images topological properties), such as object area, perimeter length etc. These properties can be used to uniquely determine the position, orientation and identity of a component in the camera field of view with respect to known database of object models. In practice this technique can be applied to a database of simple industrial components, to identify them by the silhouettes of their stable states. Various specialized lighting techniques may be applied to obtain a good image of this nature, including backlighting (e.g. an overhead projector), laser striping, and color filtering (e.g. a background painted with fluorescent red paint and illuminated with UV light increases the contrast between the background and any non fluorescent object lying on it; a red filter further enhances the contrast). In general we must define a set of geometric features which uniquely identify each object. For example, it is possible to use the following set of features which are orientation...
View Full Document

This document was uploaded on 03/12/2014 for the course MECHANICAL 214 at University of Manchester.

Ask a homework question - tutors are online