For the problem of effective fusion of observation information and target location for the large-scale swarm, the existing classical location methods either depend on the high performance or the communication bandwidth of sensors, so these methods are difficult to be applied directly to the large swarm. In this paper, a target location method based on swarm probability fusion is proposed, which can represent the sensors’ data in the large-scale swarm as a probability distribution map, and then carry out image recognition processing after normalized distributed fusion to obtain target information. At the same time, the main influencing factors of this method, such as the single node’s direction-finding accuracy, the maximum observed angle, the number of nodes, the mesh’s granularity and the relative speed, were analyzed respectively, and corresponding verification tests were carried out.