This PDF file contains the front matter associated with SPIE Proceedings Volume 10258, including the Title Page, Copyright information, Table of Contents, and Conference Committee listing.

The use of machine vision on the factory floor has not progressed as rapidly as predicted. The hurdle has been to install systems robust enough for the task, yet cost effective. With recent reductions in the price of machine vision hardware, the critical factor has been the cost of system integration. Building a solid infrastructure to support the application engineer is required to reduce integration cost. Establishing industrial standards for machine vision is a key foundation to building of this infrastructure. This paper discusses how current and future standards play a role in this building process.

The state of the art for light sources and sensors is reviewed. Lighting and viewing methods are analyzed with respect to discriminating between different types of shape and reflectance features. Techniques for viewing three-dimensional objects, including structured lighting and Moire methods, are reviewed.

Engineers can no longer confine their systems thinking to the relationships between complex technical parts. They must broaden their understanding to include people as critical elements in any integrated system. This understanding leads to unfamiliar questions, unfamiliar partnerships, and unfamiliar challenges. Operating from the broader perspective pushes engineers to expand their task and redefine their role. Including people in the definition, description, and design of new systems is essential for success in our future world.

In the course of market research carried out recurrently in Europe since 1984 it has become increasingly apparent that the growth of the machine vision industry has been hampered by a shortage of suitably skilled systems integrators. The reasons for this shortage relate both to the technology and the economics of vision related business.

There are encouraging signs indicating that both areas of difficulty are being overcome, but the prognosis is not good for the "black box’ vision manufacturer.

Many countries in the world, e.g. Germany and Japan, depend on high export rates. It is therefore necessary for them to strive for a high degree of quality in the products and processes exported. The example of Japan shows in a significant manner that a competitor should not be feared just because he can offer cheaper products. They become a “source of danger” when these products also achieve a high degree of quality. Thus, survival in the market depends on the ability to recognize the implications of technical and economic developments, to draw the perhaps unpopular conclusions for production, and to make the right decisions. This particularly applies to measurement and inspection equipment for quality control. Here, besides electro-optical sensors in general, image processing systems play an important role because they can emulate the conventional form of visual inspection by a human operator — i.e., the methods used in industry when dealing with quality inspection and control. In combination with precision indexing tables and industrial robots, image processing systems can be extended to new fields of application.

The great awareness of the potential applications of vision and image processing systems has led to a variety of realized applications, some of which will be described below under three topics:

• electro-optical measurement systems,

• automation of visual inspection tasks, and

• robot guidance.

In the past, colour vision systems have often been complex and expensive. Recent hardware and software developments have greatly simplified the implementation of colour vision systems.

At the same time, costs have fallen so that colour inspection is now practical in many industrial applications. This article examines colour vision systems in general and introduces a new approach to low cost colour processing.

This approach is illustrated by looking at a specific colour inspection task.

Now that image processing techniques applied to industrial inspection have become well established, vision systems are being used in manufacturing in increasing numbers.

This paper examines various approaches to multiple vision systems and introduces the concept of a distributed vision system.

This concept is illustrated by describing how a number of compact inspection modules can be connected to a single supervisory computer to form a distributed vision network.

The paper attempts to provide a critical overview of automated optical inspection of printed wiring boards. Some general observations of this industrial application of machine vision have been made summarizing the various approaches, the advances that have taken place and the problems that remain to be solved. A variety of both technical and trade literature is examined.

There will never be complete agreement on what constitutes the best user interface for machine vision systems. Tastes differ, after all. Furthermore, what is best will certainly depend on the application, the type of user fa- which it is intended, and the degree of training of the user. In spite of this, certain general principles can be enunciated. Without pretending to offer the “right” or “best” answers, this paper will attempt to apply these principles to user interfaces for

algorithm development systems and

systems intended for automated visual inspection in factories.

Various existing and proposed user interfaces are considered. The present “programming bottleneck” in the use of real time image processing hardware will be examined in detail.

The cost of commissioning and installing a machine vision system is almost always dominated by that of designing it. Indeed, the cost of design and the shortage of skilled vision systems engineers are together likely to be two of the most important factors limiting the future adoption of this technology by manufacturing industry. The article describes several software tools that have been developed for making the design process easier, cheaper and faster. These include:

(a) An extension of Prolog, called Prolog+. This is intended for prototyping intelligent image processing, as well as for programming future target systems.

(b) A knowledge-based program intended to assist an engineer to select a suitable lighting and image acquisition sub-system. This called a Lighting Advisor.

(c) A knowledge-based program which advises an engineer on how to select a suitable lens. This called a Lens Advisor.

(d) A knowledge-based program which assists an engineer to choose a suitable camera. This called a Camera Advisor. Ideally, items (b) to (d) should be integrated with Prolog+, so that a programmer has access to all of them in one unified working environment. Prolog+ is able to accept simple natural language descriptions (i.e., in a simple sub-set of English) of the objects/scenes that are to be inspected and is able to generate a recognition program automatically.

A range of inspection tasks is described, in which Automated Visual Inspection has, to date, made no real impact. Amongst these is the inspection of products that are made in very small quantities. An electro-mechanical arrangement, called a Flexible Inspection Cell, is described. This is intended to provide a “general purpose” inspection facility for small-batch artifacts. Such a cell is controlled using Prolog+.