Automatic Target Recognition
State of the art AI for S&R, border control and infrastructure protection
The problem
✦Most drones incorporate one or several cameras on board
✦Monitoring and surveillance are amongst the most common and potentially profitable applications of drones
✦Exploiting cameras onboard for these applications is inherently challenging
✦Manually analyzing and searching for relevant information in this video is extremally costly, or even unfeasible
✦Unlike computers, human analysis is prone to error due to fatigue and distraction
The solution
ATR
ATR is a solution for automatically detecting and classifying objects from RGB and IR onboard cameras and provide real time customized alarms at the edge, in the cloud or on-premises. ATR brings state of the art AI to S&R, border control and infrastructure protection
Get video
Get any type of video stream, in any format, from any sensor in any drone
Get telemetry
Use the drone and camera telemetry if available
Analyze
Automatically analyze video in real time
Detect
Detect, classify, track and geo-locate targets and events of interest
Get alarms
Report alarms with relevant information
ATR Benefits
Cost Reduction and Efficiency:
Reduce costs of inspection while improving efficiency, by quickly identifying targets
Consistency:
Maintain consisten performance under any conditions
Speed and Accuracy:
Analyse faster than humans for more accurate decision making
Scalability:
Scale from one drone to a fleet, with any number of video streams
Enhanced Security:
Incorporate rich information about threats and events of interest into your mission flow
Specs
Input Sources and Format:
IP Cameras; video files; video streams (RTP, RTSP, RTMP), video platforms (Milicast, Dolby.io, DJI).
Output Video Format:
Dolby, WebRTC,
RTMP, RTSP,
HLS
Integration:
REST API,
Sense Web Front
Alarms:
JSON over web socket,
Webhooks,
Serial for onboard deployments
Supported Camera Types:
RGB,
IR
Supported Targets:
Vehicles,
Persons,
Animals
Training Database:
+150K images
Over 2M annotations
Latency:
<1 ms
Deployment:
Edge,
Cloud,
Private Cloud,
Local
Mission Insight Reports
Summarize and enhance detection activity
for the full mission or selected timeframe
Interactive reports:
With the option to download as PDF
Offline and in-mission
Ideal for operators who prefer reviewing detections
over monitoring live video
Get first and last appearance
for each detected target
Includes geo-positioning data
for every detected track
Automatically display the best crop
of each target
Extra Target information
Add richer context to each detection
with advanced attributes, such as vehicle color and model
Extract license plate numbers
including partial matches
Enable alerts for specific conditions
like a red sedan matching a plate
Filter and customize alerts
based on user-defined criteria
Works in real time and post-mission
Get precise, actionable notifications
Application Scenarios
For more information, Visit our youtube channel
API Documentation
For more information about cloud integration, visit our API Documentation or send us an email
