Investigation of Machine Vision With Longwave Infrared Cameras, 10-9099

Principal Investigators
Michael P. Rigney
Ernest A. Franke

Inclusive Dates: 10/01/98 - 04/01/00

Background - Machine vision includes the technologies of acquiring images using electronic cameras and processing the images with computer algorithms. Automated inspection of manufactured parts and processes is a significant machine vision application area. Many applications have not been possible due to limitations of conventional visible light cameras. Visible light cameras cannot detect internal defects in opaque materials and have great difficulty in detecting surface defects on black rubber parts. Recent developments in infrared image-sensing technology have resulted in the availability of rugged and relatively low-cost infrared cameras, making their use economically feasible for on-line inspection applications.

Approach - The goal of this project is to develop techniques for thermal infrared inspection using machine vision. Infrared images in the 8- to 12-micrometer range were obtained from a commercial uncooled, focal plane array camera. Initial tests were conducted to characterize camera sensitivity and its nonlinear response to large temperature gradients. Thermocouple temperature measurements were used to calibrate camera sensitivity. Test samples were fabricated, and manufactured parts containing defects were obtained for investigation. Software was developed to control thermal excitation sources and acquire thermocouple temperature data synchronously with thermal images.

Accomplishments - Different thermal excitation procedures have been used to induce thermal gradients in parts under test. These procedures include radiant heating by high-intensity lamps, flash lamp, and laser; inductive heating (of metal parts and inserts); forced convection heating and cooling; and conductive heating and cooling. Thermal images and temporal image sequences of a variety of parts containing different types of defects have been acquired. Image analysis software has been developed to measure temperature contrast and heating and cooling rates from the thermal images. These signatures are used to distinguish between sound regions and defects. Investigated parts include forged and cast metal parts; injection molded plastic and rubber parts; and laminated rubber and metal and rubber parts. Developed approaches have been used to successfully detect defects that cannot be identified using conventional imaging methods.

Intelligent Systems, Advanced Computer and
Electronic Technology, and Automation Program
2000 IR&D Home SwRI Home