Solid state image scanners are based on the generation of a pattern of charge carriers in silicon crystal when illuminated. An array of small photodiodes or depletion regions measure the carrier density, and thus the illumination, at each point. By allowing the charge carriers to collect on the photodiodes or depletion regions for a defined period, the intensity is integrated over this period. The resultant charge pattern is 'scanned', i.e. read out point-by-point at one output terminal, by one of three methods: SSPA, CID, or CCD. SSPA (Self-Scanned Photodiode Array) uses an MOST multiplexer switch driven by an MOST shift register to connect each photodiode in turn onto the output terminal. CID (Charge Injection Device) injects the charge collected on an array of depletion regions into the silicon crystal, point-by-point, by switching off the field producing the depletion. CCD (Charge Coupled Device) shifts the charge pattern from point to point across the silicon surface, between depletion regions, until it reaches the output terminal.

The industrial uses of solid-state image sensors, such as photodiode, CCD and CID arrays, are reviewed. The case studies to be presented in these proceedings are supplemented by examples of and reference to further significant applications in the areas to be covered. Brief descriptions are also given of applications in other major areas. Some general techniques of optics and signal processing which are used when building arrays into instrumentation systems are also described.

The application to automated surface inspection of solid-state imaging devices, such as photodiode arrays and charge coupled devices (CCD), is discussed in relation to the other forms of electro-optical inspection systems, and the advantages and disadvantages highlighted. A specific application is described, in which a camera containing a 512 element photodiode array has been used to inspect the width and surface quality of a narrow metal strip used in the manufacture of submarine cable.

Solid state image scanners, using self scanned photodiode or CCD arrays have been used in the glass industry for a number of years and are now becoming accepted as production aids, as inspection devices, and in R & D. This paper describes the application of these devices to:- (i) The inspection of flat glass for discrete faults. (ii) The detection of windscreens broken during processing. (iii) The measurement of optical distortion produced by windscreens. (iv) Dimensional measurements on hot glass during processing. (v) The measurement of displacement during the dynamic strength testing of glass.

The measurement of lateral track parameters, using contacting sensors, from a high-speed rail vehicle presents problems due to the vulnerability of mechanical probes and feelers. To overcome these problems a non-contacting system based on commercially available solid-state scanners has been developed by British Railways. The sensor equipment consists of two separate linescan cameras mounted vertically above each rail. The rails are obliquely illuminated from projectors mounted on the bogie (truck), and it is possible by processing the camera video signals to detect -that part of the scan corresponding to the gauge face of each rai1. This information can then be used to calculate track gauge, and, in conjunction with other equipment, track alignment.

A solid-state image scanning technique is described which allows the diametral straining of directly heated, internally pressurized tubes to be monitored continuously up to the occurrence of rupture. The system utilises a 512-element photodiode array coupled to a processor unit which provides a BCD output to a paper-tape punch at determined time intervals; this tape is analysed by a mini-computer to give a graphical record of the test. Examples are given of the application of the mechanical properties data, obtained from the strain-time plots, to predictions of deformation behaviour relevant to the performance of fuel tubes in nuclear reactors.

Many food manufacturers face the problem of product which is sold by weight having to fit into packages of fixed volume. This is particularly the case when making biscuits which are sold in the familiar roll-packs. Automatic measurements of product dimensions at intermediate stages of the process assist operators in keeping the product within specification. No commercially available device met the requirements for a non-contacting measuring instrument at reasonable cost but a simple adaptation of IPL's Linescan camera system provided the answer. In the device developed, a well-defined beam is projected at an angle onto the surface to be measured, and the displacement of the light spot as seen from vertically above is detected using the Linescan camera.

The authors will describe a technique initially developed for automatic inspection and gauging of extruded products which has more recently been applied to on-line volume measure-ment of bulk materials carried on conveyor belts. The technique involves a light-sectioning optical system in which a straight edge of light is projected across the direction of flow. When viewed from a different angle, this edge is seen to follow the external profile of the material, and this view is imaged onto a two-dimensional photodiode array. Processing of the array output signal with an electronic system including a microprocessor enables the instantaneous cross-sectional area of the load to be computed, since the underside profile formed by the empty conveyor belt is known. Integration of these measurements over a given length of conveyor movement yields a measurement which is proportional to volume.

Inspection of manufactured components for automatic assembly identifies two distinct aspects of the problem. The more obvious requirement of checking for tolerance is clearly important. However, a separate and equally important feature of the problem is the examination of components for gross defects which arise during manufacture. It is these which lead to the 'jamming' of feeders prior to the assembly operation. The work described in this paper details methods of checking for gross defects within a microprocessing environment. It is anticipated that the cost is such that it could be tolerated within the budget for a parts feeding package. The flexibility of integrating a linear photo-diode array with a microprocessing computer provides a further, and perhaps more significant, advantage in that the determination of orientation of component parts is inherent within the system. It is conceivable therefore, that a general 'computer' tooled bowl feeder might prove a real competitor to the mechanically tooled alternative.

When a worn or damaged cutting tool is replaced on a numerically-controlled machine tool, the control system must be informed of the coordinates of the cutting edge on the new tool with respect to the tool holder. An automatic setting system is desirable, particularly if tool wear is rapid. In the instrument to be described, a silhouette of the tool tip is imaged onto two linear self-scanned photodiode arrays which effectively form a crosswire fixed with respect to the machine bed. The tool is driven towards the crosswire position and when its image intercepts the arrays the control system stores the coordinates of the tool holder.

NEL was recently required to produce two servo-controlled goniometers for a crystal x-ray monochromator. The specification included an angular sensitivity and repeatability of 0.036 arc second and a linearity better than 0.1 arc second. Operational requirements dictated both an incremental system for operation around an experimental datum and an absolute system for fixed system datum operation, the latter enabling spatial synchronism after a switch-off. A three-track optical grating was chosen for angular pick-off, with a digitizer for coarser measurements. The finest grating track consisted of 36 000 lines at a pitch (grating wave-length) of 36 arc seconds: to approach the required sensitivity an interpolation factor of 1000 was necessary. Conventionally, moire patterns formed by a grating/index interface are detected by four photo-cells in a dc manner and the produced output is related to the light intensity, mean modulated light transmission, modulation amplitude and moire angle. Drift can be experienced to an extent dependent upon system type, and the voltage dependent interpolation methods are limited to factors of 10-40 for confident interpolation. Replacement of the dc photocells by a dynamic self-scanned photocell array enables modern signal processing techniques to be applied to the resultant ac temporal phase signal. This enhances the measurement quality, increases the interpolation possibilities by two orders, and gives immunity from all reasonable dc drift. Some measure of compensation for defects in the optical system can also be achieved. Four such ac moire detectors were fitted to the goniometers to provide a measure of spatial averaging, and single detectors to the coarser tracks for absolute build-up. The systems meet the specification and are presently undergoing field trials.