Abstract: In order to improve component temperature retrieval precision, the component ratio of pixel was acquired by reflectance spectrum data of ASTER(advanced spaceborne thermal emission and reflection radiometer). Because the histogram method was introduced to the genetic algorithm, the optical ranges of parameters research were reduced. And the improved genetic algorithm was used to retrieve three component temperatures: vegetation, sunlit soil and shade soil temperature in the YingKe study area and three temperatures were simulated. To provide a good test data, the retrieval of component temperatures by genetic algorithm under condition of only considering thermal radiation was validated. The results showed that comparing simultaneous field data, the error range of simulated temperature under condition of considering thermal radiation and reflectance data was 0.763%-8.867%, and the error range of simulated temperature under condition of only considering thermal radiation was 8.955%-11.832%, the study results show that the new method is feasible and provide a new way of thinking for retrieval of land surface component temperature.