With the widespread application of nuclear technology, the risk of nuclear accidents such as nuclear leakage, nuclide proliferation, and loss of strong radioactive sources has increased. Therefore, it is necessary to properly dispose of radioactive pollutants to avoid social panic and casualties. However, due to the limitation of strong ionizing radiation in the nuclear environment, for emergency tasks in the nuclear environment, using mobile robots to search for radioactive sources in the environment has become a new development trend. Also, due to the influence of search time and detection performance of radiation detectors, mobile robots can usually only obtain limited radiation measurement data. However, to facilitate radiation safety monitoring and subsequent radiation source disposal, it is necessary to obtain the dose rate distribution map and the location of the radiation source in the environment. To solve the above problems, this paper designs a mobile robot search strategy to collect sparse radiation measurement data, inverts the regional radiation field based on Gaussian Process Regression (GPR) method, and then uses Hough Transform (HT) method to locate unknown radioactive sources. Experimental results show that, in single-source localization experiments, the error remained within a small range and gradually decreased to approximately 0.16 m as the sampling data increased. In multi-source localization experiments, the highest mean error was 0.91 m, while the lowest was 0.15 m.