In order to enhance the operational performance of non-sensitive projectiles and achieve target recognition as well as precise attack, a YOLO-1D model based on the idea of YOLO model positioning is put forward in this paper to realize target identification and location of non-sensitive projectiles composite detection signals. Based on the signal variation characteristics of laser and infrared detection signals when a target is present, labels are designed and a dataset is constructed in line with the YOLO concept. Meanwhile, a network structure named Multi-Task Convolutional Neural Network (MT-CNN) based on multi-task learning strategy, and a random forest recognition algorithm relying on artificial feature extraction are constructed to serve as control group. Finally, the target recognition and localization performance of the above models are evaluated and compared with the UAV detection test data. The test results show that the YOLO-1D model achieved the best performance in terms of recognition and localization., and its recognition accuracy rate reaches 95.61% and the positioning accuracy rate reaches 81.10%.