A two-stage method for constructing real-time high-precision temperature map

摘要

Temperature monitoring has many real-life applications in the fields of electricity, meteorology, manufacturing, agriculture, security, etc. Therefore, it is important to efficiently construct a high-precision temperature map based on a number of sampling temperature data collected by temperature sensors. For this problem, the traditional methods include the average value method, the finite element method, and the Kriging method, among which the Kriging method has the highest accuracy, but requires the highest complexity, preventing its applications in large-scale scenes. To overcome this drawback, a two-stage constructing method is developed, which divides all the pixels into two classes, called 1-level pixels and 2-level pixels respectively. For the 1-level pixels, it still uses the Kriging method to estimate the corresponding temperatures (to ensure accuracy). Based on estimated temperatures of the 1-level pixels, for the 2-level pixels it uses a local weighted mean method with low complexity to estimate the corresponding temperatures (to improve efficiency). Experimental results show that, the new method could greatly improve the efficiency of calculation, only with very few loss of accuracy, thus satisfying the real-time requirement in large-scale cases.