Laser radars produce three-dimensional pictures

Imaging radar systems, capable of providing range information over an extended two-dimensional (2-D) field of view, have significant advantages over conventional cameras. While 2-D technology can provide clues about the relative distance of objects within a scene, there are many cases in which this information is ambiguous or deceiving. In a classic example, determining the relative distance of two vehicles that are camouflaged to blend into a particular background is difficult or impossible using intensity methods. In contrast, imaging laser radars are not only capable of picking out the camouflaged objects but can unambiguously determine their distance from the transmitter.

Because of their high cost, systems of this type have traditionally been used for defense and research applications. However, this is slowly changing as imaging radars have found their way into certain commercial applications in which accurate three-dimensional (3-D) imaging provides significant economic benefit. These applications are expected to grow in number during the next decade as technological advances lower the cost of imaging radar units. This cost reduction likely will be fueled by the availability of lower-cost components and the development of new system architectures. In the case of "staring" systems, good examples of new designs incorporating components originally designed for the home-entertainment market already exist.

Conventional ranging techniques

Laser radar systems of all types determine the distance from the transmitter to an object (ranging) by measuring the time required for light to travel to the target and return. The hardware required to accomplish this task is, to first order, dependent on the detection scheme used. In point-detecting radars, there are two major detection schemes (see Fig. 1).¹ In one, light from an amplitude-modulated transmitter is collected by the receiving telescope and directly converted to an electrical signal by a suitable detector. Commonly referred to as energy- or direct-detection radars, systems of this type have relatively straightforward hardware that places few restrictions on the spectral properties of the laser transmitter.

FIGURE 1. Laser-radar information can be obtained via two techniques. In direct or energy detection (top), the electronic signal leaving the detector is proportional to the amplitude of the return. In heterodyne systems (bottom), the return from a single-frequency transmitter is mixed with a stable local oscillator generating a difference-frequency output.

The other major design is based on a heterodyne detection scheme requiring a single-frequency transmitter and local oscillator. In this architecture, the target is illuminated by the amplified output of a single-frequency local oscillator or a single-frequency pulsed laser. Reflected light is collected by a telescope and optically mixed with a local oscillator whose single-mode output is frequency-locked to the transmitter. This process produces a detector output signal at the difference frequency between the return beam and the local oscillator. Frequency modulation of the transmitter and/or Doppler frequency shifts produced by movement of the target can be determined and used to calculate range and/or target velocity.

While systems using either type of detection can measure the location of a point object with a high degree of accuracy, they provide little information in the plane perpendicular to the line of sight. Such information can, however, be generated by scanning the angle of the transmitter output and collecting data at a number of points in the scan cycle. In many cases, the viewing angle of the receiver is also scanned by placing a moving aperture between the receiving telescope and the detector. For example, one scanning system that has been developed for industrial inspection uses direct detection and a sinusoidally modulated diode-laser transmitter to provide unambiguous range data that are accurate to 1/10 in. for objects located 1 to 2.8 m from the transmitter.² In this system, range data are collected at regular intervals as the transmitter and receiver are scan ned and subsequently assembled into a 3-D image by a computer.

While scanning systems have been under development for many years, progress in the development of fast detector arrays has recently led to the demonstration of high-resolution staring systems in which data for the entire image are collected at once. One of the most promising systems of this type is currently being developed at Sandia National Laboratories (Albuquerque, NM).³ With this system, the target is flood-illuminated by the transmitter output. At the receiver, range data are generated across the entire viewing aperture by modulating the gain of a microchannel-plate image intensifier placed in front of a standard CCD array. Using a 512 × 512-element detector, readout rates of 15 Hz have been reported for this system, which is capable of "inch-type spatial and radar range resolution at distances out to 1 km." The images at the top of this page were taken using this approach.

Applications of imaging radars

Imaging laser radars, like many other laser-based technologies, offer unique capabilities at comparatively high cost. Both NASA and the military are interested in the technology for applications that range from the docking of autonomous spacecraft to obstacle avoidance, bathymetry, and smart munitions. Initially developed under the sponsorship of these agencies, systems are slowly modified for other uses and, with sufficient cost reduction, are introduced into the commercial marketplace.

While the $50,000 to $100,000 price tag of currently available systems has slowed this transition, certain applications exist in which the speed and resolution of an imaging radar device result in process improvements that more than offset the cost. One example is the inspection of the oxygen furnaces used by the steel industry. Approximately 60 feet in length and three feet high, these furnaces are operated continuously and are turned off only for maintenance. During operation, the integrity of the refractory lining is of major concern because an unexpected failure results in a major personnel safety problem and significant system downtime. Scanning laser radars have been used for this application, and staring systems are currently being evaluated.

In another interesting application, diode-laser-based scanning radars are used in the forestry industry.² Modern lumber mills are highly sophisticated facilities that use the current price data on finished lumber in combination with a 3-D map of a log to determine the optimal cutting configuration. Due to the high throughput of such a mill, variations in yield on the order of 1% can result in significant savings. When compared to 2-D imaging and laser triangulation, laser radar offers an attractive combination of speed and accurate 3-D mapping. Using multiple radar units in a configuration similar to that shown in Fig. 2, a computer-generated picture of the log is constructed, and the optimal cutting pattern determined. When compared to 2-D imaging and laser triangulation, the radar produces a better image at higher speed and can easily achieve the 1/10-in. resolution required by the application.

FIGURE 2. Multiple diode-laser-based imaging radar systems are used to generate three-dimensional images of uncut logs in production lumber mills. Subsequent computer analysis uses data on the current cost of finished lumber to determine how the log is ultimately cut.

Cost reduction needed

While niche markets for imaging radars will continue to develop, significantly larger, untapped markets exist for lower-cost units in the fields of machine vision and inspection. For example, a scanned imaging radar can be used in concert with a specially designed "picking" algorithm to help a robot select and grab a component from a bin of parts (see Fig. 3). In an initial demonstration, a Perceptron (Plymouth, MI) system was used to guide a single-armed robot as it picked exhaust manifolds from a bin. By using an algorithm that identified all possible points at which the robot could pick up parts within the bin, an average selection time of 15 seconds was achieved.

FIGURE 3. Although the current cost is somewhat high for many applications, imaging laser-radar systems can provide the high-resolution, 3-D images needed for robotic control. In this figure, a single radar located over a box of exhaust manifolds allows closed-loop computer control of a picking operation in which the robot selects, grabs, and places individual pieces on the table to the right of the bin.

Unfortunately, many applications of this type require a lower-cost radar, the development of which will require progress on several different fronts. In the case of scanning systems, much of the cost of the unit lies in the scanner, and a significant cost reduction will require the development of a low-cost unit that combines high speed and accuracy. In addition, smaller cost reductions are possible for both the laser transmitter and processing electronics.

With staring systems, the cost of the scanner is eliminated, but a more powerful (and costly) laser is required in addition to a sophisticated receiver. In the case of the Sandia system described above, the cost of a microchannel-plate image intensifier and high-resolution CCD array is currently in the $5K to $10K range, with the laser transmitter and processing electronics adding additional cost.

Fortunately, this cost has been minimized by incorporating computer-based image-processing and rendering technology developed for the consumer market. In a recent version of the system, 3-D visualization and rendering is accomplished in real time using a Pentium-based PC and a low-cost video accelerator card. On the laser end, reductions in the cost of high-power laser-diode arrays, principally driven by the requirement of the diode-pumped solid-state laser industry, should lead to a slow reduction in source cost.

It is reasonable to expect a continued increase in the performance, reliability, and speed of imaging laser radars, driven primarily by the needs of the defense and research sectors. In some cases, this process will result in new system architectures that can be marketed at reduced cost. More likely, however, significant reductions will occur slowly as the cost of individual components (lasers, scanners, detector arrays and processors) drops. As demonstrated by the Sandia system, this reduction will be the greatest in those cases for which it is possible to use components developed for large consumer markets such as computer gaming and digital image processing. Unfortunately, such markets have yet to develop for lasers, scanners, and image intensifiers, and the availability of an imaging laser radar at less than $10K is probably some years in the future.

ACKNOWLEDGMENTS

The author wishes to thank John Sackos of Sandia National Laboratories (Albuquerque, NM), Gary Kammerman of Fastmetrix (Huntsville, AL), and L. van Wezel of Perceptron (Plymouth, MI) for their input during the preparation of this article.

REFERENCES

1. J. W. Goodman, IEEE Transactions on Aerospace and Electronic Systems, AES-2, 526 (1966).

2. H. K Roberts and L. van Wezel, SPIE Proc. 2748, 2 (1996).

3. J. T. Sackos et al., SPIE Proc. 2748, 47 (1996).