Abstract: The compaction quality of RCC construction is very important to dam forming quality. The existing compaction quality detection method of RCC layer is nuclear density instrument method, which has the disadvantages of complex operation, low reliability, safety risk, regular calibration and the lack of representativeness, and cannot meet the requirements of rapid and accurate detection. Through the theoretical and experimental research, the material parameters of the field compaction layers, that is, the moisture content of the mixing material, the gradation of aggregate and the propagation speed of the stress wave of the compaction layer were selected as the evaluation parameters in this paper, and the corresponding real-time warehouse surface moisture content tester and the roller thermal layer wave velocity tester were developed to collect and transmit the above parameters in real-time. And then a prediction and evaluation model for compaction of RCC based on moisture content, gradation of aggregate and propagation velocity of stress wave was established by using BP neural network. Through the remote visual feedback system, a whole set of real-time compaction feedback and control technology of thermal layer in RCC construction was formed. This technique was applied in the construction site of the Wunonglong RCC dam, and the construction effect of the RCC thermal layer was monitored in real time.