VisionTracer-

Video Investigation Big Data Platform


Product Introduction

Based on visual recognition calculation and ReID cross-lens tracking technology, the VisionTracer video reconnaissance big data platform can accurately locate and quickly query targets through centralized processing, analysis, classification, research and judgment and dynamic management of images. The system supports real-time control, multi-clue trajectory depiction, target behavior analysis and other functions



Functional Characteristics

undefined


RelD Cross-camera Tracking

Intelligent recognition of targets such as pedestrians, motor vehicles, and cyclists, based on the fine-grained attention mechanism neural network algorithm (RA-CNN), extracts multi-dimensional feature information.


The target is continuously tracked and identified across domains and lenses, and its trajectory is tracked, recorded, and restored for analysis.

undefined


Accumulate A Large Number of Actual Business Samples

Based on a large number of samples of Skynet, Xueliang, Electric Police bayonet and Social Monitoring, the algorithm has been trained in multiple rounds of iteration to highly match the actual business scenarios.


It can train in multiple scenarios such as multi-person, multi-vehicle, multi-target, and multi-season to expand the adaptability of the algorithm.


Automatically extract

pedestrian features

Feature extraction of the human body.


Identifythe same person acrossmultiple lenses

Under different cameras, different time periods, different seasonal conditions, and different clothing, the target can still be located.


Share clues   

Overlay multiple tracks.

Support multi-person collaboration, when the team members find the clues, the clues can be shared, and the multi-person clues can be superimposed on the multi-person trajectory map.


Display of multi-person trajectory diagrams

On the map, the trajectories of multiple people in different places and at different times are superimposed.

Application Scenarios