With the aim to deal with the localization disadvantage of robot equipped with only single class sensor, a novel mobile robot particle filter self-localization method based on combination of the heterogeneous sensors was proposed. Perception model of LRF (laser range finder sensor) and monocular camera were established, and self-localization was achieved after the particle sets had been updated with fusion perception information. The experimental results showed that characteristics of fast and accurate updates of LRF and global of monocular camera was fully utilized, convergence time of particle sets was reduced by 14.3% than using a single class of sensor, and mobile robot located accuracy was improved by 16.7%.