| 摘要 |
Method for optimising processing of images, from image sources (eg. CCTV cameras), with overlapping areas, comprising: generating a 2D map representing a real scene; partitioning the part of the map representing the overlapping portion into multiple sub-areas using an occlusion criterion; associating a camera identifier with points in the overlapping region based on the occlusion criterion or on a detection confidence score; generating a role map with multiple parts based on the map and identifiers, each part of the role map being associated with a single camera identifier so that only one camera is selected to process each portion of the overlapping area. The purpose of the role map is to identify and select which camera to use to process each sub-region of the overlapping area, eg. some sub-portions of the overlapping area may be visible to one camera but occluded to others the optimal camera to use will be the one that is not occluded and will therefore be assigned to that sub-region. A camera may be associated with a sub-section if the confidence of object detection exceeds a threshold. Occlusion criteria may be when bounding boxes of people or tracking targets overlap at the same time. |
