Special Report: Optical Patterns
However, the approach also raises several issues to be resolved, i.e., Aberrations that are due to the oblique angles of light propagation, or off-axis holographic optical elements. Limited optical path lengths are available on a thin substrate. A high resolution lithographic technique is required for fabricating high-quality (large space-bandwidth product) diffractive optical elements to replace conventional bulk lenses in this approach. This is even more serious in this case than in conventional diffractive optical components in free-space optics because the light path length is short and all the other components need to be scaled down accordingly.
A hybrid integration of a microlaser chip and a planar optics substrate has been implemented [89]. In the system, a top-surface emitting microlaser array (9mm diameter, 85mm spacing, and 850nm wavelength) was bonded to a single quartz glass substrate by the flip-chip bonding technique with indium solder bumps. The lenses were lithographically fabricated by the diffractive optical element technique with four phase levels. They have a square shape with a size of 500mm X 500mm. Their focal lengths match the thickness of the glass substrate, which is 3 mm. Surrounding areas are coated with a Ge film to absorb any undesired stray light. The deflection angle relative to the normal is limited by the resolution of the lithographic process; it was 6.2o in this experiment. Various other approaches along similar lines have been proposed and implemented [91-93].
12.7 Conclusion
Optical pattern recognition has often been superseded by the rapidly growing digital computer technology that approaches near-real-time processing. However, it should be noted that the computational power of optics to compute 2-D Fourier transformation, convolution, and correlation at the speed of light is still unique and is not expected to be surpassed by other technologies. Even with current device technologies, optical pattern recognition can achieve more than 1,000 image comparisons per second. Extrapolating the developmental speed of digital computers doubling every two years, 1000 high resolution image comparisons per second (about two orders of magnitude faster than what is available now) would require another 15 years or so and will not be attained by the year 2010. Before then, the explosion of information in the form of multimedia through the internet will increasingly demand faster and faster image comparisons and understanding, for which optical pattern recognition can provide many opportunities. Fortunately photonic technology forms the backbone of the current communication technology. The current transmission speed of an optical signal is moving beyond 1 THz, much faster than the clock speed of a digital computer. Total available optical bandwidths can be as high as 25 THz at the communication wavelengths. These bandwidths give a clear edge to optics-based information technology for storage, displays, and processing. In this regard, optical pattern recognition has a bright future if it is closely tied to other photonic technologies and takes advantage of new photonic devices. For example, the developmental speed of SLM's has been much slower than that of photonic devices for communications. Undoubtedly, surface emitting microlasers will have rich new possibilities for pattern recognition as the device technology becomes mature enough to allow low (of the order of microwatts) power consumption per laser, high reliability, integration with other devices, and suitable addressing schemes for high density arrays, etc. Current optical pattern recognition systems need to be greatly improved in terms of not just speed but of accuracy. This accuracy requires the development of elaborate measurement methods to calibrate and test performances. Such test and measurement issues need to be seriously addressed in the future before any optical systems are to be used for practical purposes.
Previous Next Table
of Content for report: Optical Patterns Home
