Skip to main content

Detection Filters

Effective filtering is essential for optimal Lookout+ performance. This guide covers filtering strategies for different operational scenarios and system configurations.

Understanding Detection Filters​

Detection filtering interface
Detection filtering controls for optimizing display and alerts

Filtering controls what detections are displayed and processed, directly affecting:

  • Visual clutter on detection displays
  • Alert frequency and relevance
  • System performance and processing load
  • Integration with autonomous systems like GAMA

Existence Probability Filtering​

Threshold Strategy​

The existence probability threshold determines which detections are shown based on AI confidence:

Lower Thresholds (30-60%):

  • Show more detections, including uncertain ones
  • Higher false alarm rate
  • Better for human monitoring where false alarms can be ignored
  • Useful for research and comprehensive surveillance

Higher Thresholds (70-90%):

  • Show only confident detections
  • Lower false alarm rate
  • Better for automated systems and collision avoidance
  • Reduces processing load and alert fatigue

Application-Specific Settings​

GAMA Integration:

  • Recommended: 75-85% threshold
  • Rationale: Prevents autonomous systems from avoiding non-existent objects
  • Impact: Reduces unnecessary course changes and false collision alerts

Human Monitoring:

  • Recommended: 50-70% threshold
  • Rationale: Operators can visually verify uncertain detections
  • Impact: Maintains situational awareness while reducing obvious false alarms

Marine Mammal Observation:

  • Recommended: 40-60% threshold
  • Rationale: Early detection allows time for avoidance maneuvers
  • Impact: Captures more potential wildlife encounters for research

Environmental Filtering Strategies​

Clear Weather Conditions​

  • Lower thresholds: Take advantage of optimal detection conditions
  • All object types: Enable comprehensive monitoring
  • Fine-tune by time: Adjust for changing light conditions

Poor Visibility (Fog, Rain, Spray)​

  • Higher thresholds: Reduce false alarms from weather effects
  • Focus on larger objects: Prioritize vessel detection over small objects
  • Use thermal cameras: Switch to IR detection when available

High Sea States​

  • Increased thresholds: Minimize false detections from wave action
  • Stability considerations: Account for camera motion affecting detection quality
  • Size filtering: Focus on objects larger than typical wave features

Night Operations​

  • Thermal camera priority: Rely on IR detection for primary monitoring
  • Adjusted sensitivity: Account for different thermal contrast conditions
  • Light source awareness: Consider artificial lighting effects on detection

Quick Reference​

ScenarioThresholdFocusNotes
Autonomous Navigation80%Vessels, large mammalsReduce false avoidance
Collision Avoidance70%All vessels, hazardsBalance sensitivity vs false alarms
Marine Mammal Research50%Mammals, swimmersHigh sensitivity for early detection
Harbor Operations75%Vessels, navigation markersReduce clutter from small objects
Open Ocean60%All typesComprehensive monitoring