As shown in Table 1, the time-of-flight system offers the best combination of system performance, cost, and ease of use for many applications. Stereo visionís ability to use standard high-resolution cameras makes it the best system for 3D movie creation. Structured lightís ability to scale its accuracy by implementing additional complex light patterns makes it the best system for very high precision scanning applications. For any application that values low latency, precise gesture tracking, or medium precision scanning, time-of-flight is ideal.
Table 1. A summary of the strengths and weaknesses for these three systems.
(Source: Texas Instruments)
Regardless of which method is utilized, once the system has the depth information, the data flow from that point forward is the same. There are two levels of software required in a 3D imaging system. The middleware analyzes the depth map to identify objects and patterns and extract the critical data points needed for the end-application. For example, in a gesture application, the middleware identifies the major body parts (hands, head, feet, fingers, etc.) and builds a skeleton of data points used for tracking. In a scanning application, the data to be extracted includes points of inflection on the objectís surface. The application layer takes the tracking data extracted by the middleware and performs the function desired by the user: move a tool into place, start/stop a process, rotate an object for inspection, and so on.
HMI in industrial automation
As more and more complex systems are being utilized in the manufacturing environment, operator safety and ease-of-use are becoming hot topics. As operators can be required to manage multiple systems at once, the ability to control them with gestures rather than touch can make their job simpler and safer. Now they can keep their eyes on the actual moving parts of the equipment/process, rather than the control interface. This can help them to identify problems earlier and shutdown the process quickly from where they are, rather than moving to a control system and push the correct button, or flip a switch or lever. Imagine how fast it would be to stop the machine by simply holding up two closed fists or making an X with your arms crossed in front. Or imagine moving a precision tool into the ideal location with precise finger motions. This could be in three dimensions while at an optimum viewing angle, versus utilizing a joystick on a control unit. This could replace moving it manually, which could expose the operator to potential injury from the tool itself.
Robotics vision has been around for quite a while now. Imagine how much more accurate the robotís movements can be if it has true 3D information available, rather than a simple 2D image. The time-of-flight system's low latency and lack of precise mechanical alignment requirements make it ideal for such systems. The precise alignment required for either stereo vision or structured light make them very difficult to mount on a robotic system that typically moves at high speeds and is subjected to repeated mechanical shock and vibration. The result in either of those two systems would be getting knocked out of alignment, decreasing the performance and requiring more system downtime for maintenance and recalibration. Furthermore, the low latency of time-of-flight is perfect for enabling more rapid motion and decision making. The Texas Instruments time-of-flight system is capable of supporting frame rates of up to 120 frames per sec (fps). This enables depth information to be generated at rates of at least four times that of competitive solutions.
Scanning and inspection
As discussed above, structured light is a great system for very precise scanning systems. However, when less precise scanning is required, as is the case in many manufacturing environments that want to automate the final "go/no-go" decision, a time-of-flight system is more than adequate. Its precision and low latency enables faster decisions to be made and increases manufacturing throughput. Regardless of methodology chosen, a scanning system may require the ability to rotate either the object being inspected, or the imaging system around the object. The mechanical requirements of structured light implementations potentially can limit their applicability in factory environments that subject the systems to extreme shock and vibration.
Factory automation can leverage 3D imaging to enhance system performance and ensure operator safety. Gesture control can be used to allow the operator to stay focused on the process, rather than the control system, and function as a more natural machine interface. Machine vision that understands depth as well as X-Y coordinates can enable robotic systems with precision movement. Utilizing 3D-scanning systems, more accurate object measurements and analysis is possible. Finally, 3D time-of-flight systems, like those offered by Texas Instruments, offer capabilities never before imagined in factory automation and machine control.
Dan Harmon is sensing business development manager for TI's sensing group.