UAV Depth Map Processing With Multithreading for Obstacle Detection

The patent divides the image processing workflow into three independent modules: image correction, binocular matching, and obstacle distribution acquisition. Each module is executed by a separate thread, allowing parallel processing of image data. This segmentation enables the system to maintain high frame rates without processing bottlenecks, as multiple operations occur simultaneously rather than sequentially.

The patent introduces ring queues as intermediary data structures between processing threads. The ring queues serve as buffers that decouple the production of image data from its processing, allowing threads to operate independently without blocking each other. This mediator mechanism enables smooth data flow between modules while maintaining parallel execution, thus preventing processing delays even at high frame rates.

The patent segments the processing workflow into distinct modular components that can be executed in parallel. By separating image correction, binocular matching, and obstacle detection into independent threads, the system achieves faster execution without requiring complex inter-dependent processing logic. Each module remains relatively simple while the overall system benefits from parallelization.

The patent performs preliminary actions by pre-configuring multiple threads and ring queues before processing begins. The thread pool and buffer structures are established in advance, allowing immediate parallel processing once data arrives. This preliminary setup eliminates the need for dynamic thread creation during processing, reducing overhead while maintaining simplicity.