This paper presents an intelligent multisensor system for 3D metrology tasks, which combines 2D intensity data coming from a CCD-camera and 2D range data delivered by a light stripe triangulation sensor. Due to the high data-rates both of the image processing system as well as the range sensor the multisensor system can perform the 3D-inspection of moving parts, making the system well suited for manufacturing lines based on closed loop sensing and control technology. The basic approach described herein makes use of intensity data for guiding the light stripe sensor in X- and Y-direction to areas of interest on the specimens surface. Guidance information is derived from a symbolic scene description created during a three stage segmentation algorithm. The segmentation algorithm consists of edge detection, Kalman-filter based contour tracking and contour approximation.

An image acquisition and processing algorithm for inspection of tire treads has been developed. The tire treads are flat strips of black rubber material used as the main component in retreading automobile tires. These treads have a complex molded design on one side (DESIGN SIDE) and a flat surface on the other side. The inspection of the Design Side of the tread is one of the key operations in the tread fabrication process impacting quality and consistency of the final product. This paper will discuss development of the main optical inspection algorithms utilized in the system design. The algorithms described in this paper were tested in the laboratory prototype of the inspection system and will be implemented in the final production system.

An integrated inspection system for improving the accuracy of CNC machine tools is proposed. The system described in this paper emphasizes the integration of the on-machine inspection and analysis techniques with the information coming from post-process and on- machine inspection to improve machine performance automatically. Algorithms are derived for analyzing the post-process and on-machine inspection data to identify residual systematic errors and relate them to the machine performance. Various data analysis algorithms and techniques are compared. A feature comparison approach is developed to relate the dimensional and form errors of a manufactured workpiece to the systematic machine tool errors. Inverse kinematics technique and statistical methods are used to identify and characterize the contribution of each geometric error component. A self tuning algorithm is also proposed to fine tune the geometric-thermal model.

A machine vision system was developed to inspect roses, and assign grades similar to those used by human graders. Illumination and image processing techniques were developed to extract the main features used by human vision. These features were identified as stem length, diameter and straightness, and bud color and maturity (i.e., openness). Illumination techniques include the design of an inspection chamber and the choice of proper light source. Image analysis includes the development of algorithms for image understanding. Supervised learning was used to classify roses.

This paper presents a series of unique methods to generate a complete B-Reps type solid model, based on the information of stereo image pairs from multiple views that are obtained by taking pictures of polyhedral objects with stereo-cameras. First, we present a method to generate 2D B-Reps that become a basis of topological information of 3D B-Reps, and a hierarchical method to determine corresponding point pairs. Then, by introducing an open shell, a set of visible faces with a boundary, we present a method to generate the open shell regularly in terms of geometry to represent a result of recognition for a single view. Finally, after discussing the problems when unifying plural open shells obtained for multiple views into a solid, we propose a method to overcome them and generate a complete solid (a closed shell) model. This paper asserts that by outputting the B-Reps in terms of the data structure of a 3D CAD a variety of applications become possible, and that, this model, which generates open shells in an intermediate stage, is also effective in reducing shape errors (relative positional errors).

A new programmable 3D laser range camera is developed for Robot Teleoperation in hazardous environment. The aim was to build a 3D laser range camera with attractive features for users friendly Teleoperation of mobile robot. BERTIN & Cie was in charge of the development of the 3D laser imager and ETNOTEAM was in charge of the development on an user friendly Man-Machine Interface for Teleoperation purpose. The current presentation describes in details the new characteristics and performance of the 3D Laser Range Camera. Furthermore, preliminary results of the system in an environment including shining metal objects are proved.

A fully automatic crosscut saw, based on automatic detection of all relevant wood defects, gives completely new opportunities to determine the optimum use of a wooden board. In this paper, a quality model is given to match the requirements of the end products to the quality of the raw material. The quality model is well suited for efficient computer implementation. Furthermore, an optimization procedure based on this model is proposed, giving an integrated system where defect data is input in one end and final cutting lists are output in the other. The procedure is guaranteed to produce optimal cutting lists from the given input data. Tests with real data from a Swedish window manufacturer show that the average output value is 7% higher using this procedure than what is obtained from their existing commercial saw. The procedure is also very efficient. The test shows that it is capable of producing optimal cutting lists for their given product mix within an average 17 ms on an ordinary PC.

In this paper, we present the development of a fast and robust vision system which, in conjunction with the Dial Gage Calibration system developed by AKO Inc., will be used by the U.S. Army in calibrating dial gage torque wrenches. The vision system detects the change in the angular position of the dial pointer in a dial gage. The angular change is proportional to the applied torque. The input to the system is a sequence of images of the torque wrench dial gage taken at different dial pointer positions. The system then reports the angular difference between the different positions. The primary components of this vision system include modules for image acquisition, linear feature extraction and angle measurements. For each of these modules, several techniques were evaluated and the most applicable one was selected. This system has numerous other applications like vision systems to read and calibrate analog instruments.

Infrared thermography allows the instant passive thermal mapping of a remote surface, noninvasively and without contact. Under controlled conditions precision temperature measurements can be made of these surfaces. IR thermographic equipment has evolved to the point where it is portable, versatile, rugged and reliable; applications are found in virtually every phase of manufacturing and materials inspection. This paper traces the evolution of IR thermography from its beginnings to the present. Current commercially available equipment is classified and described and an overview of applications is presented. Finally several specific examples of industrial applications are described and illustrated, and a glimpse of the near future is offered based on the most recent developments in equipment and software.

It has been shown that given any polygonal part, it is possible to derive a sequence of grasping operations that are sufficient to orient the part up to symmetry without sensors. At the conclusion of such a sequence, one degree of translational freedom is part position remains. This paper describes how to eliminate that degree of freedom with a sequence of pushing motions. Thus we show that sensors are not necessary to uniquely determine the final position and orientation of polygonal parts.

Assembly planning based on force-based reasoning of decomposable subassemblies, clusters of parts that are naturally broken apart from an assembly by decomposition forces is presented. The proposed force-based approach does not depend on combinatorial cut-set enumerations, but it generates only the minimal cut-sets of force-flow networks through a max-flow/min-cut algorithm, which incurs only polynomial time complexity. The key features of the proposed approach are as follows: (1) force-flow networks are constructed from the liaison graph of an assembly to represent the maximum amount of forces that individual parts can deliver to their mating parts for the chosen disassembly directions, fixtured and grasped parts, and removable connectors, (2) decomposable subassemblies are identified from the minimal cut-sets of force- flow networks based on a max-flow/min-cut algorithm, (3) the assemblability of the identified decomposable subassemblies are evaluated based on a force-based stability analysis, and (4) planning is done by the recursive decomposition of an assembly into subassemblies and verification of its reversibility.

In this paper we discuss the approach to model a real-time control system for a two part type, two machine, two buffer manufacturing system with finite set-up change times for machines. The environment is stochastic due to random machine breakdown and repair times. A four level hierarchical structure is used for this purpose. The objective is to minimize the work-in- process inventory while satisfying the demand. The demand rate is considered to be piece-wise constant. Simulations are made on the system to study its performance.

Image analysis can be used to characterize granular populations in many processes in food industry or in agricultural engineering. Either global or individual parameters can be extracted from the image. However, granular products may appear agglomerate on the image, bringing biasing on individual parameters. Combining statistical and neural network technics enables the build of a system which can recognize if products are agglomerate or not. To process images after agglomerates detection, two approaches have been studied: the first is based on erosion, followed by conditional dilation with the original image; the second takes advantage of the graph's properties of the agglomerate's skeleton.

The paper deals with the subject of robot calibration. With particular reference to assembly robots, a brief introduction of the mathematical approach to the problem is presented. It is shown that two independent type of measurements can be carried out to obtain data for calibration. A first optical set-up is used to measure robot gripper orientations along straight trajectories. The system is based on laser triangulation. A second set-up is used to measure X,Y,Z position of the robot gripper programmed to follow a number of proper straight trajectories. The paper also presents some experimental results from the calibration of a commercial SCARA robot.

Previous generation robots were originally designed for lead-through on-line programming and were found to be quite successful in that mode. During the past decade, some of these robots were retrofitted with additional software to perform straight line interpolation between program points taught using a teach pendant. Given this new `cartesian' model, a natural enhancement was to then use such robots in a simulation and off-line programming environment. However, the results of initial tests showed that while such robot systems was repeatable to within a few millimeters during playback of previously taught on-line programs, they could be inaccurate by as much as six inches when programmed off-line. This paper discusses the quantification of positioning errors for such an application using automatic coordinate measuring theodolites, the identification of the robot kinematic signature describing the differences of the actual robot kinematics from the nominal drawings, and the compensation of the geometric and non-geometric robot errors by modification of the robot software model and translator in a workstation based graphical off-line programming system.

For multi-robot coordination experiments, traditional algorithmic approaches through partial optimization methods have given incomplete solutions with very heavy computational burdens. Complete neural-fuzzy solutions are still in exploratory stages, and are yet to be worked out. This paper addresses problems of planning coordinated robot motions by applying 3D measurement of human arm motions. This paper then proposes an improved fuzzy expert controller which executes optimally-planned paths given by the coordinated-motion planner. This fuzzy controller consists of a rule-based fuzzy tuner and intermittent algorithmic optimizer.

This paper describes the development and use of the Multi-Axis Seam Tracking (MAST) sensor for tracking seams or other features in real-time. Four independent, spatially- distributed electric fields are used to sense changes in the relative position of the sensor and the workpiece. The MAST sensor is very inexpensive compared with commercially available seam tracking sensors. It can be used in system to perform cost-effective small-lot manufacturing operations in a faster, more consistent manner. The MAST sensor is used in an automated system for dispensing braze paste during a rocket nozzle fabrication process.

This paper discusses the problem of tracking a geometric object using data from a real-time laser rangefinder. Because of the volume of data we choose to preprocess each range image to find the edge points. Then 3D lines and circles are extracted from these edge points. By matching these extracted curves to the model of each object we are able to compute the object pose, that is its position and orientation. Some examples are given of the operation of these algorithms using simulated range data.

In this article, two methods performing fine assemblies of micromechanical parts are presented. The first one is an optical method based on projection moire technique and associated with a phase shifting procedure. The second one uses tactile sense principle with a high accuracy robotized site and a force sensor equipped gripper. The aim of this paper is to compare these two methods. Finally, we give some interesting observations about their results which can help in using them in a complementary and efficient way for assembly.