Pedestrian target detection needs to process a large amount of image or video data,which puts forward high requirements on the computing power and storage capacity of AI chip in terms of real-time performance and efficiency.Therefore,the localization AI chip pedestrian target detection technology with multi-label and multi-target fusion is proposed.The feature extraction model of pedestrian target is constructed,and the feature information is spliced across dimensions.The parallel computing capability of AI chip is used to fuse the above features.The convergent characteristics of the target space are obtained by combining the rectification function.Based on this,considering multi-target and multi-fusion,the IOU threshold is used to calculate the probability of pedestrian target candidate frame,so as to achieve high-precision pedestrian target detection.The experiment results show that the proposed method can detect pedestrian targets with high accuracy.