|
Machine vision began in the 1950s with pattern recognition of two-dimensional images, initially designed to replace human eyes for inspection and identification work, significantly improving inspection efficiency and reducing inconsistencies in results caused by eye fatigue. Machine vision inspection has evolved to the point where it can accomplish tasks difficult for human eyes, such as high-precision measurements and high-speed grading of specific products, as well as utilizing infrared, ultraviolet, X-ray and other detection technologies to detect things invisible to human vision.
Machine vision systems are making progress in many areas, including deep learning software, 3D vision, and modularity that supports process flexibility and user choice. With the introduction of deep learning software in machine vision systems, capable of detecting unpredictable defects simultaneously, it can identify anomalies on high-speed production lines within milliseconds and learn from small sets of sample images within minutes. Deep learning will enable machines to develop their own product specifications and solve previously unsolvable problems, such as distinguishing between stains on cell phones and scratches on casings. In addition to appearance and functional defect detection, the software can classify textures and materials, verify assembly, locate deformed parts, and read characters, including distorted printing and optical character recognition (OCR) text. The latest deep learning software can also analyze classified datasets, such as images labeled by human operators, to program itself based on representations of good or bad parts and other factors, making applications previously beyond machine vision capabilities possible.
As machine vision systems grow in speed and power, 3D machine vision is transitioning from controlled industrial environments to unstructured industrial environments.
Modular vision allows customers to choose what type of embedded lighting and optical elements they want to use in their smart cameras, or add barcode scanning and basic machine vision functions to standard devices like smartphones.
Combining vision with software and automated production systems enables the greater flexibility needed for today's manufacturing and packaging of e-commerce products. Vision systems can distinguish colors or identify text on shiny surfaces, while new lighting systems will adapt to product changes by automatically adjusting wavelength, angle, and height. All these factors make it easier to meet flexible manufacturing needs. Vision systems need to adapt to many variations in products and materials, as well as detect contaminants or tears—the earlier defective products are found, the less material waste occurs.
Traceability and record-keeping supported by vision systems are becoming increasingly important for combating mass counterfeiting in industries such as automotive, pharmaceuticals, medical devices, and components. New systems are using inks invisible to the human eye to print codes that can be hidden in any color or image, even under paint, with these identification codes detectable through special functions. Traceability is also crucial for the increasing use of medical devices implanted in patients.
