Abstract: Current methods for retrieving surface temperature using remote sensing data and point data from ground temperature sensor networks yield low temperature inversion precision. To solve this problem, collaborative inversion methods with ground temperature sensor network(GSN) data and remote sensing inversion data fusion were explored four solutions for combination ground sensor network technology and remote sensing based on HJ-1 data, which were proposed to retrieve ground temperature. Experimental results shown that root mean square error of four solutions respectively decreased from 0.8848℃ to 0.6562℃, 0.4288℃, 0.4535℃ and 0.4261℃, and the correlation coefficients increased from the initial 0.6195 to 0.6343, 0.8629, 0.8507 and 0.8629. Moreover, the temperature error of solution four was below 0.45℃ and correlation coefficients were above 0.85 in the case of increasing pixel intervals. The results were validated using different images and GSN data. A comparison of the results and analysis of the models shown that the new model combining brightness temperature with classification results increased the accuracy of the initial retrieved results.