Autonomous Driving Object Recognition Device Adaptive Calculation Mode

The patent applies local quality by differentiating calculation modes for different objects based on their priority levels. High-priority objects (those with higher collision risk) are processed in high calculation mode to maintain recognition accuracy, while low-priority objects are processed in low calculation mode to reduce overall system load. This selective approach ensures critical objects receive full processing resources while preventing system overload.

The system dynamically changes the calculation mode parameter based on real-time assessment of object priority, frame complexity, and resource occupation state. By adjusting the calculation mode (high, low, or skip) as a variable parameter rather than maintaining a fixed high calculation mode, the system achieves both accurate recognition of critical objects and overall system stability.

The patent applies local quality by assessing each object's priority independently within complex frames and assigning appropriate calculation modes. This allows the system to detect and track all objects (maintaining quantity) while consuming fewer computational resources by processing only high-priority objects in high calculation mode and low-priority objects in low calculation mode.

The system applies partial action by selectively providing high calculation mode processing only to the extent necessary for high-priority objects, rather than uniformly applying high calculation to all objects. This partial intensive processing approach maintains comprehensive object detection while preventing excessive resource consumption that would lead to system overload.