Vehicle Surround-View Imaging for Glass and Reflective Object Positioning

The patent segments the image processing by separating light-transmissive and reflective objects from ordinary objects. It uses specialized processing paths: ordinary objects are mapped using conventional space model mapping, while light-transmissive and reflective objects are detected through polarization characteristics and infrared radiation, then positioned using stereo vision techniques. This segmentation allows each object type to be handled with appropriate methods, resolving the positioning accuracy problem.

The patent changes the detection parameters for different object types. For light-transmissive objects, it uses polarization degree parameters to distinguish them from ordinary objects. For reflective objects, it uses infrared radiation characteristics. These parameter changes enable accurate detection and positioning that conventional methods miss, directly improving measurement precision for these specific object types.

The patent applies partial action by not processing all captured images through the full three-dimensional mapping pipeline. Instead, it uses a hybrid approach: conventional space model mapping for most areas, and selective stereo vision processing only for regions containing light-transmissive or reflective objects. This reduces calculation complexity while maintaining information quality where it matters most.

The patent introduces an intermediary detection step using polarization cameras and infrared sensors to identify objects requiring three-dimensional processing. This intermediary layer filters which regions need complex calculation, allowing the system to maintain high information quality for critical objects while avoiding unnecessary calculations for ordinary scenes, thus balancing complexity and information quality.