In the path planning of target search based on probabilistic maps, the target is usually set to obey discrete uniform distribution in the working space, and then the length of the path is employed as an index to optimize the global path of the search task. However, most of the probability distributions of targets in the real working environment do not satisfy uniform distribution, which will lead to the search path not being the one with the shortest expected time. In order to solve this problem, a probability calculation model is designed according to actual working environments, and based on which, a probability map is designed. With the probability map, an expected-time optimal target search path planning method for robots is proposed, in which, a sequence planning method is used in the upper topology map to obtain the observation point search sequence of the optimal expected time. Finally, a local path planning method is used in the lower feature map to obtain the collision-free path between the observation points. The experimental results show that this method can significantly shorten the expected time of target search, and is more suitable for working situation where the target does not obey uniform distribution.