Abstract: A reliable method of grain storage supervision can effectively guarantee the quantity of grain storage in and out of warehouses according to plan and reduce the loss of unplanned entry and exit. In recent years, there has been a method of monitoring grain storage by video equipment, but the security of storage and management of video surveillance is poor and inconvenient to use. In this paper, we proposed a method for grain storage supervision based on the similarity of RGB color features of temperature field cloud map. Firstly, the historical grain data of the grain storage was called and pre-processed to remove the random code, error and other data. According to the correlation of the temperature at the adjacent temperature measurement points, the grain temperature data of each plane in the grain bulk was interpolated and the temperature field cloud map was generated. Then the similarity of the temperature field cloud map at the adjacent days was calculated by the similarity algorithm based on the RGB color feature distribution, similarity threshold was set according to the similarity of cloud maps during normal storage. Finally, the abnormal movement in the grain bulk was judged according to the similarity threshold. In order to verify the feasibility of grain storage regulation based on similarity of temperature field cloud map, five kinds of abnormal movement in grain bulk were simulated. The five kinds of abnormal movement respectively were: the half part of the grain bulk at right side and latter side, the quarter part of the grain bulk at right side and the latter side, and overall of the grain bulk. Similarity algorithm based on the RGB color feature distribution was used to detect abnormal movement of grain bulk. Meanwhile, the method based on the similarity of LBP texture feature was also used to compare with the method, the results showed that the mean of recall rate of the method based on the RGB color feature distribution was 98.6%, the mean of precision rate was 97.3%, and the operation speed was about 320 ms/time. The mean of recall rate of the similarity detection algorithm based on the LBP texture feature was 97.3%, the mean of precision rate was 96.2% and the operation speed was about 540 ms/time. The data were analyzed by analysis of variance, the results showed that the influence of anomaly types and temperature plane on recall rate was very significant and the influence of abnormal type on precision rate was very significant, and the influence of temperature measurement plane on precision rate was not significant. Taking into precision rate, recall rate and algorithm speed consideration, similarity detection algorithm of cloud map based on RGB color feature distribution was more suitable for the detection of grain storage supervision. The test of grain storage supervision was carried out, and the results showed that the algorithm can not only regulate the grain storage, but also detect the local heat in the grain bulk. The purpose of this study was to lay the foundation for a reliable and simple regulatory approach to grain storage regulation.