Multihop range-free localization method is an effective and simple method to determine node location. However, it is generally suitable for isotropic and the node intensive network. Since the low performance of multihop range-free localization method in anisotropic networks, an improved scheme is proposed based on canonical correlation. This method obtains a high precision mapping model by canonical correlation regression between nodes hops and Euclidean distance, then obtains the estimated distance between unknown node and known node using the measurement of hop-count. It also solves the problem of multi-hop range free localization method, which the performance is low in anisotropic network. Experimental simulation results show that this method achieves higher accuracy and stability than other location algorithms.