Occupancy Grid Object Sizing With Edge Confidence Tracking

The patent segments the object detection process into multiple independent components: occupancy grid generation from sensor data, cluster algorithm for cell assignment, prediction equations for state estimation, and Kalman filter for temporal tracking. Each component processes specific aspects separately, then integrates results to achieve both high precision and robustness.

The patent implements feedback through the Kalman filter which continuously compares predicted object states with actual sensor measurements, calculates prediction errors, and adjusts subsequent predictions. This closed-loop feedback mechanism enhances both measurement precision and detection robustness by correcting errors over time.

The complex processing is segmented into distinct modules: occupancy grid creation, clustering algorithm for cell assignment, prediction equation calculation, and Kalman filter processing. This segmentation allows each module to be optimized independently while maintaining overall system manageability and processing efficiency.

The patent performs preliminary actions by pre-calculating prediction equations based on object motion models and pre-establishing the occupancy grid structure before actual detection. This preparation reduces real-time computational complexity while maintaining high precision in object state determination.

The patent applies periodic iterative processing through the Kalman filter at optimized time intervals rather than continuous processing. The filter iterates at discrete time steps, updating object states periodically based on new sensor data, which maintains tracking robustness while reducing overall processing time compared to continuous iteration.

The patent changes processing parameters dynamically by adjusting the iteration frequency and prediction horizons based on object velocity, distance, and detection confidence. High-priority objects with uncertain states receive more frequent iterative updates, while stable, well-tracked objects use less frequent updates, optimizing the balance between robustness and processing time.