Fujitsu Laboratories said its technology is the first of its kind that can detect people from low-resolution imagery in which faces are indistinguishable.
Detecting the movements of people could be useful for retail design, reducing pedestrian congestion in crowded urban areas or improving evacuation routes for emergencies, it said.
Fujitsu used computer-vision algorithms to analyze the imagery and identify the rough shapes, such as heads and torsos, that remain even if the image is heavily pixelated. The system can pick out multiple people in a frame, even if they overlap.
Using multiple camera sources, it can then determine if two given targets are the same person by focusing on the distinctive colors of a person's clothing.
An indoor test of the system was able to track the paths of 80 percent of test subjects, according to the company. Further details of the trial were not immediately available.
"The technology could be used by a business owner when planning the layout of their next restaurant/shop," a Fujitsu spokesman said via email. "It would also be used by the operators of a large sporting event during times of heavy foot traffic."
People-tracking know-how has raised privacy concerns in Japan. Last year, the National Institute of Information and Communications Technology (NICT) was forced to delay and scale down a large, long-term face-recognition study it was planning to carry out at Osaka Station, one of the country's busiest rail hubs.
The Fujitsu research is being presented to a conference of the Information Processing Society of Japan being held at Tohoku University in northern Japan. The company hopes to improve the accuracy of the system with an aim to commercializing it in the year ending March 31, 2016.
Fujitsu has also been developing retail-oriented technology such as sensors that follow a person's gaze as he or she looks over merchandise as well as LED lights that can beam product information for smartphones.
Tim Hornyak covers Japan and emerging technologies for The IDG News Service. Follow Tim on Twitter at @robotopia.