Aiming at the problems of complex network structure and inaccurate target depth information obtained after depth estimation in monocular 3D detection, this paper proposes an end-to-end joint multi-attention depth estimation monocular 3D target detection network structure (CDCN-3D). Firstly, in order to obtain the salient features of the target, the adaptive spatial attention mechanism is introduced to aggregate the pixel features to enhance the local features and improve the network representation ability; Secondly, in order to improve the problem of local information loss in depth estimation, the improved C-ASPP is used to capture more accurate direction perception and position sensitive information for each depth information. Finally, the accurate P-BEV is used to map the three-dimensional information of the target to a two-dimensional plane, and then the single-stage target detector is used to complete the detection and output task. The experimental results show that the accuracy of CDCN-3D network is better than other monocular 3D detection networks on KITTI data set. In Car, Pedestrian and Cyclist classes, its detection accuracy is increased by 2.31%, 1.48% and 1.14% respectively, and it can complete the 3D target detection task.