Though OMRON's endeavor to develop Silhouette Vision technology has a relatively short history, the road to its development has been a long one with many twists and turns. Overcoming these difficulties one by one eventually led to the successful implementation of Silhouette Vision technology.

Using two cameras to emulate human eyes is not as easy as it first appears. For instance, to ascertain the distance from a viewed object, it is first necessary to know the amount of space between, and the viewing angle of, each camera "eye." Because people process this information unconsciously, they can estimate the distance to an object or its height as soon as they see it.

Distance measurement mechanism of human eyes

For a machine to perform this same function as humans do, it must be first informed of the interval between the two cameras and each camera's viewing direction. In addition, the technical challenge of processing immense quantities of image data as intelligently and speedily as possible must be overcome. To meet these challenges required the best of OMRON's accumulated expertise in sensing technology, and the result was a truly compact and low-cost Silhouette Vision technology-based sensor. This in turn led to the successful commercialization in 2000 of a traffic flow sensor using two cameras, the first of its kind in the world. Even today, OMRON remains the only company that offers traffic flow sensors based on two cameras.

The commercialization of a Silhouette Vision technology-based traffic flow sensor presented many hurdles that were not anticipated at first. One difficulty was creating a sensor that can demonstrate reliable performance regardless of environmental conditions. Because these sensors must be installed outdoors on the road, vehicle images taken by cameras may vary greatly according to weather conditions, which can change every day between sun, rain or storms. In regions with heavy snowfall, many cars on the road may have snow piled on the roof, which also had to be taken into account.

Difficulties were not limited to weather factors. As the sensor was destined for global markets, it was essential to ensure that it could correctly respond to situations outside of Japan. For example, in the United States, the very size of vehicles differs greatly from those in Japan. On US highways, it is not unusual to see trucks of 30 meters or longer. To make sure that the sensor functions appropriately in varied conditions overseas, OMRON's development team actually brought the sensor to the installation sites and repeated on-site checks and adjustments. By overcoming these hurdles one by one, OMRON finally perfected a high-performance Silhouette Vision Sensor with a recognition accuracy of over 97%.

Around 1997, OMRON launched a development project for Silhouette Vision using two cameras. The idea of distance measurement based on triangulation principles (connecting two eyes) is not a new one, and some technical papers describing ways to implement the principles into machines had already been published at that time. However, hardware that could put the idea into practice was not available until the late 1990s. It also coincided with the time when information technology began to show dynamic advancement, spurring on the rapid dissemination of the Internet among the general public. The performance of personal computers improved rapidly, and PCs with unimaginably high performance and memory capacity became accessible at amazingly low costs. Along with these factors came declining prices for the cameras themselves. These performance and cost factors enabled OMRON to devise a mechanism for compressing image data using two cameras. The mechanism allowed image data to be processed at an exceptionally high speed-almost in real time-leading to the successful market launch of the Silhouette Vision Sensor.
"On the road, there were many more variants besides weather conditions that had to be taken into consideration, such as lane changes and sudden turns by vehicles, than we could imagine," says Masaki Suwa, in charge of developing the Silhouette Vision Sensor. "To respond to these challenges, we repeatedly conducted tests in various locations around the world, finally succeeding in the perfection of the Silhouette Vision Sensor that is available today. In the past, a camera's main function was recording, but we believe that it will become increasingly important for extracting information from a recorded image in much the same way that humans see. Toward the goal of creating the best match between machines and people, we are determined to develop smarter sensors."

Masatoshi Kimachi, who is in charge of Vision Sensing Technologies at OMRON's Sensing Technology Laboratory, also expressed his enthusiastic determination to further promote the research and development of Silhouette Vision. He remarked, "Many different systems are currently in operation throughout our society. To effectively control them, the capacity of sensors is the most important factor. Our Silhouette Vision has tremendous potential. We are striving to use this technology for industrial machinery, such as in parts sensing, or to build it into vehicles and home-use robots in the future. Sensing using two cameras offers almost unlimited possibilities for new applications."

Return to Vision Sensing Top