基于多模态图像融合的俯拍目标检测的方法、设备及介质

本申请公开了一种基于多模态图像融合的俯拍目标检测的方法、设备及介质，涉及图像识别技术领域。方法包括：获取待检测区域的俯拍彩色图像以及与所述俯拍彩色图像一一对应的深度图像；将所述俯拍彩色图像、对应的所述深度图像进行图像对齐；对对齐后的所述深度图像进行兴趣目标区域过滤，得到第一目标深度图像；将所述俯拍彩色图像融合到对应的所述第一目标深度图像；通过形态学操作以及鬼影消除处理对融合后的所述第一目标深度图像进行背景过滤，得到第二目标深度图像；基于所述第二目标深度图像对应的点云数据和所述俯拍彩色图像进行目标检测，得到若干头肩目标。设备及介质应用上述方法，本申请的实施例能提升目标检测的效果。

The invention discloses a high-angle shot target detection method and device based on multi-modal image fusion and a medium, and relates to the technical field of image recognition. The method comprises the following steps: acquiring high-angle-shot color images of a to-be-detected area and depth images in one-to-one correspondence with the high-angle-shot color images; performing image alignment on the high-angle shot color image and the corresponding depth image; performing interest target area filtering on the aligned depth image to obtain a first target depth image; fusing the high-angle shot color image to the corresponding first target depth image; performing background filtering on the fused first target depth image through morphological operation and ghosting elimination processing to obtain a second target depth image; and performing target detection based on the point cloud data corresponding to the second target depth image and the high-angle shot color image to obtain a plurality of head and shoulder targets. The method is applied to the equipment and the medium. The target detection effect can be improved.