When Proctor Penetrometer is used to measure the compaction degree of filled soil in field, the variation of soil water content would bring about large error to the measurement results. In this article, a new method for compaction measurement integrating Proctor Penetrometer and soil moisture sensor is proposed and the soil models are built based on dual-sensor data fusion to reduce the influence of water content on the compaction test. Compacted loess samples with different water content and dry density are prepared. The penetration resistance measured by Proctor Penetrometer and the output voltage measured by soil moisture sensor are compared with the dry density determined accurately by sand replacement method and the water content determined accurately by oven-drying method. Binary regression analysis method is applied to conduct data fusion and set up a mathematical model involving
dry density, penetration resistance and output voltage. Test result indicates that after fusing the penetration resistance with the output voltage, the dry density can be measured directly and less influenced by water content and closer to the true soil dry density value, thus the influence of water content on the measurement results is greatly reduced.