Occlusion Object Detection via Collaborative Sensing Deep Convolution Network

Object detection is one of the important problems in computer vision. But external occlusion often cause object features missing which lead to a big challenge of object detection. Aim at the problem of occlusion object detection and try to describe object features more effectively; we proposed a collaborative sensing deep convolution network to achieve co-detection by global and partial features of objects. Firstly, we divide the global and partial of the object, it means we segment parent and child in an object. Then, the joint detection network of parent and child is constructed. Finally, through the collaborative detection we achieve the precise positioning and recognition about parents. The proposed algorithm effectively solves the problem that object can not be detected due to missing features. We also ensure the accuracy of parent construction by child. Experiment results demonstrate that our algorithm performs better than other state-of-the-art methods.