As one of the key technologies of unmanned vehicles, target detection is prone to missed detections and false detections due to occlusion or incomplete targets. at the same time, the detection accuracy is low. To address this issue, this paper proposes a novel improved YOLOv3 network for real-time high-accuracy pedestrian incomplete object detection. First, on the basis of Darknet53, three attention modules(SENet, CBAM, and ECA) are integrated, and the number of network layers is simplified to build a new network structure with small parameters and high precision, which improves the recognition ability of the algorithm. Secondly, at the neck of YOLOv3, a spatial pyramid convolution pooling module is used to enhance the deep feature extraction capability of the network at a minimal computational cost. Finally, the CloU loss function is used as the bounding box regression loss function of the network, which improves the detection accuracy of the network for the target. Experimental results demonstrate the effectiveness of our network in pedestrian incomplete object detection.