How can a parked vehicle monitor its surroundings without keeping its main processor active? A low-power AI chip handles detection at the camera edge.
OMNIVISION has introduced the OAX7700, an ultra-low-power AI companion processor designed to add always-on security functions to automotive camera systems.
Parking surveillance requires cameras to remain active even when a vehicle is stationary, but continuously running a central computing platform can increase power consumption. The OAX7700 addresses this by handling initial detection locally through an integrated neural processing unit (NPU).
The chip can process camera data to detect people and objects, estimate distance, and identify potential events while the system remains in a low-power monitoring mode. When an event is detected, the system can switch to normal operation and retain five to 10 seconds of pre-event video for further analysis by the vehicle’s electronic control unit.
This approach can help distinguish an actual security event from movements caused by birds, trees, wind or debris. OMNIVISION says systems using the OAX7700 can consume up to 97% less power than existing parking surveillance solutions.
The edge processor also allows manufacturers to develop their own AI algorithms for functions beyond security. Software updates can be deployed over the air, allowing those functions to be modified without changing the camera hardware.
The OAX7700 measures 5.5 × 3.5 mm and combines stacked PSRAM and Flash memory in a 60-pin BGA package. Samples are available, with mass production scheduled for the first quarter of 2027.
“With the OAX7700 ASIC, we are enabling existing automotive camera solutions with highly efficient parking surveillance functionality while at the same time supporting modern software-defined vehicle architectures,” says David Ho, senior marketing manager at OMNIVISION.
For the official announcement, click here.
