Object detection in images
The present invention provides a conditional form for segmentation maps to “grow” segmented blobs to fill holes and intersections without replaying shadow pixels with segmentation to create an object detection method and system that is resistant to lighting changes that cause shadows and / or highlights.
Abstract
The present invention applies a segmentation operation to the input image to identify the foreground object of interest, and then applies a shadow removal operation to remove any shadows detected from the foreground segmentation. However, the shadow removal algorithm can leave holes and intersections in the segmentation map that will subsequently affect the object detection step performed using connected component analysis. To circumvent this problem, the present invention provides a conditional form for segmentation maps to “grow” segmented blobs to fill holes and intersections without replaying shadow pixels with segmentation. Apply mathematical expansion operations. The result is an object detection method and system that is resistant to lighting changes that cause shadows and / or highlights. [Selection] Figure 2