Target-oriented grasping is one of the fundamental problems in robotic manipulation and holds significant practical value. However, grasping specific target objects in cluttered environments remains a substantial challenge. To improve grasping success rates for target objects, this paper proposes a novel Hierarchical Pushing and Grasping (HPG) strategy. HPG adopts a hierarchical network architecture, where the grasping network acts as a discriminator to guide the training of the pushing network, thereby enhancing the coordination between pushing and grasping sub-actions. Furthermore, a dedicated reward function is designed to alleviate the sparse reward problem in unstructured environments. Finally, both simulation and real-world robotic experiments are conducted to to assess the effectiveness and generalization of the proposed HPG strategy of the proposed HPG strategy.