Localized edge computing Box (ZS-AIBOX-RK3588J)

Localized edge computing Box (ZS-AIBOX-RK3588J)It integrates a variety of edge computing intelligent algorithms such as panoramic stitching, target detection, target tracking, target recognition and alarm linkage, which can be applied to different combat scenes such as air, ground and water, and has the ability to detect, recognize and track multiple targets such as people/vehicles/animals/aircraft/ships/UAVs.

Localized edge computing Box (ZS-AIBOX-RK3588J)To enable the security system to have the practical effect of listening to all directions, observing all directions, and being able to make good decisions, achieving a smooth upgrade from traditional security to intelligent security, and effectively improving monitoring efficiency.

Filter out false alarms caused by wind, grass movements, and various external environmental interference factors (such as trees, rocks, buildings, flags, railings, flying insects, rainwater, etc.), accurately detect targets such as personnel/vehicles/animals/aircraft/ships and drones, and automatically overlay and store target types in the video.

Support setting detection conditions according to specific requirements, including target wandering, retention, retrograde movement, gathering, intrusion, crossing boundaries, etc. It can mark targets in real time and upload alarm information.

By continuously tracking the target, the target is locked in a certain area of the central screen, and the tracking frame is automatically corrected throughout the tracking process. When there is an error in tracking, the target can be automatically re framed. At the same time, the lens adapts to zoom to ensure good tracking effect.

▲ Facial recognition

▲ Ship type identification

Using deep residual convolutional neural network as the basic model, through target recognition and feature extraction/comparison, the multi-dimensional matching degree of the target is calculated based on the Euclidean distance calculation method, and the target attributes are judged. For example, facial/license plate recognition, vehicle/ship classification, and plant growth period recognition.

Communication with various external devices is achieved through interfaces, enabling access to various alarm sources and coordinated handling of alarms. When an external device detects an alarm, based on the location information of the alarm, it can guide the photoelectric device to automatically jump, confirm the event twice, automatically identify and track the target, and record the entire process of the event.

▼

Support real-time splicing and output of multiple 4K video frames, with frame synchronization between each video frame, and no ghosting, misalignment, or missing content at the splicing point. Front end splicing greatly reduces the bandwidth pressure on network transmission.

Support integration with external business systems on site (such as ADS-B, AIS, GPS, etc.) to obtain multi-dimensional information about targets and display them in videos.

▲ Product appearance and interface board