A hybrid camera system built by researchers at Shibaura Institute of Technology (SIT) combines wide-angle target monitoring with high-resolution image capture. The system detects indistinct target regions during omnidirectional viewing, and then uses pan-tilt cameras to capture a high-resolution image of the target. To make the system more efficient, the researchers developed an algorithm for reducing the number of complementary shots required. The combination of 360° monitoring with high-quality imaging capabilities makes this hybrid system potentially useful for surveillance applications that require wide field of view without compromising their ability to recognize and clearly image distant objects. The platform constructed by professor Chinthaka Premachandra and researcher Masaya Tamaki comprises two pan-tilt cameras with a 180° field of view on either side, in combination with a monocular omnidirectional camera. Two fisheye lenses sandwich the body of the omnidirectional camera, with each lens covering a 180° capture range. The pan-tilt cameras allow the capture of 180° ranges on either side of the omnidirectional camera. Researchers from Shibaura Institute of Technology (SIT) designed a camera platform using an omnidirectional camera for target detection and separate cameras for high-resolution capture to allow accurate object identification without incurring large computation costs. The new camera platform could be used in security and surveillance systems. Courtesy of Chinthaka Premachandra, SIT, Japan. The pan-tilt cameras are Raspberry Pi Camera Modules v2.1 mounted on a Pan-Tilt HAT by Pimoroni Ltd., connected to and controlled by a Raspberry Pi 3 Model B (Raspberry Pi). The researchers connected the complete system — the omnidirectional camera, the pan-tilt cameras, and the Raspberry Pi — to a personal computer to allow computational control. To operate the system, the researchers first process an omnidirectional image to extract a target region. The target’s coordinate information is converted into angle information (the pan and tilt angles necessary to capture the target) and transferred to the Raspberry Pi. The Raspberry Pi controls each pan-tilt camera based on the angle information and determines whether to capture a complementary image of the target region. The researchers performed four primary experiments to demonstrate the performance of four different aspects of the camera platform. In separate experiments, they demonstrated the image-capturing performance of the platform by setting and imaging the target objects in different locations. Experimental results showed that use of this camera platform to capture target regions allows acquisition of higher-resolution images compared with using a single omnidirectional camera. The researchers recognize that a potential issue could arise when a moving object is to be captured as the target region. In this case, captured complementary images could shift due to the time delay required for image acquisition. They propose a potential countermeasure, which is to use a Kalman filtering technique to predict the future coordinates of the object when capturing images. The hybrid camera system developed at SIT could enable accurate object identification without incurring large computation costs. “We expect that our camera system will create positive impacts on future applications employing omnidirectional imaging such as robotics, security systems, and monitoring systems,” Premachandra said. The research was published in IEEE Sensors Journal (www.doi.org/10.1109/JSEN.2021.3059102).