New Camera Platform Features Wide-Angle Monitoring with High-Resolution Image Capture
This Raspberry Pi-powered, two-in-one camera system enables 360-degree monitoring for target detection and its capture in high resolution.
Researchers from the Shibaura Institute of Technology have designed a hybrid camera system capable of taking wide-angle, high-resolution images. With most cameras, there is a trade-off between the field of view and resolution, which is more prominent with omnidirectional cameras that provide a 360-degree FOV. For example, security cameras typically provide a wide-angle view of any given area, and while most objects in the foreground are visible, those further away become obscured.
The team developed a dual-camera system that employs an omnidirectional camera for target detection and a separate lens for high-resolution images to overcome those limitations. The hybrid system features an omnidirectional camera and a pan-tilt (PT) camera with an 180-degree FOV positioned on either side. The omnidirectional camera also sports a pair of fisheye lenses on either side of the camera body, each covering an 180-degree capture range. The researchers outfitted the platform with a pair of Raspberry Pi Cameras mounted on a pan/tilt module, then connected the entire camera system to a Raspberry Pi 3 Model B. The platform was tied into a computer for overall control.
The hybrid camera system works by first processing an omnidirectional image to the same target region it was taken from, with the coordinate information converted into angle information, which is passed on to the Pi. The Raspberry Pi then processes the data and determines if the cameras need to take additional images based on the resolution. The team performed a series of tests to demonstrate the platform’s performance by taking images at four different target locations, each providing a high resolution of the area.
While the system can take HD images, capturing moving objects presented an issue due to the time delay of capturing the target; however, they rectified the issue by employing the Kalman filtering techniqe, which predicts the future coordinates of an object when capturing images.