|The Image Sensor Spreading over Japan
My current responsibility is to improve the performance of our image sensor, "SECOM AX." "SECOM AX" is one of our core products aimed at detecting intruders using image recognition technology. "SECOM AX" sensors provide safety and assurance to customers 24 hours, in offices and stores all over Japan. This sensor has extreme performance criteria since its performance is related directly with our customers' safety and peace-of-mind. To achieve the required performance, many researchers in our laboratory have contributed to improvements and as a result, the "SECOM AX" sensor has become extremely precise and advanced.
Developing an image sensor which can perform well under the complexities of real-world use is a challenging research area with many researchers around the world involved in tackling this difficult problem. Although humans have the natural ability to recognize and differentiate objects that they see, many technical hurdles remain to develop an automated system to undertake this same task. Since our image sensors deployed throughout Japan must work in a wide variety of environments, the sensors that we develop must perform well in all of these environments.
A truly practical algorithm
In my work, I am focusing on developing a practical algorithm by looking at the problem from a different angle than the one traditionally taken by others in the past.
My thought process always involves first listing the events or objects to be recognized, and then analyzing their physical properties. Next, I find some intrinsic features exhibited by those physical properties. After that, I try to develop approaches to mathematically represent those intrinsic features, which enables the sensor to recognize the existence of those features. Of course, I also track algorithms devised by researchers around the world, to learn from their approaches and perhaps apply this new information to improving our image sensor.
I believe the most important thing is to make sure that the algorithm performs well by repeatedly testing in various real and computer simulated environments. This is achieved by repeating the cycle of, 1) testing, 2) analysis, 3) identifying problems, 4) generating ideas, 5) implementing the ideas, until the problems are resolved. Since the quality of the ideas is main factor affecting the performance of the algorithm, I use my knowledge and experience to their full extent to come up with good, effective ideas. Although it may take a long time and a lot of effort to develop a truly practical algorithm, this is a task that has to be done thoroughly to prevent the development of a faulty algorithm. This process is one of the most thrilling experiences that university research cannot typically offer, but researchers at SECOM enjoy often.