Abstract: In this research, a new automatic device was designed for the indoor soil column infiltration detection, which was mainly composed of a sensor position adjusting device, a soil sample laying device, a water supply device, a frequency domain reflectometry (FDR) soil moisture sensor, a CLZ-A pressure-strain type sensor, an Arduino mega 2560 development board, an Arduino UNO development board, a TB6600 byte rotor driver and upper desk software. The sensor position adjusting device was composed of the FDR soil moisture sensor, a 28BYJ-48 byte rotor, two 42BYGH47 byte rotors and four limit switches. The FDR soil moisture sensor was driven by 28BYGJ-48 byte rotor in the horizontal direction. The FDR soil moisture sensor was driven by 42BYGH47 byte rotor in the vertical direction. Two limit switches were fixed on both sides of the FDR soil moisture sensor in the horizontal direction. The other limit switch was fixed on both sides of the FDR soil moisture sensor in the vertical direction. The soil sample laying device was a cylinder with holes made of acrylic materials. The probe of FDR soil water moisture sensor could go in or out of the holes. The cylinder was driven by a 57BYG250B byte rotor until the soil moisture sensor probe could go in or go out from the holes in the cylinder. The water supply device was composed of a Markov bottle, a CLZ-A pressure-strain type sensors and a support board. The CLZ-A pressure-strain type sensor was placed in the bottom of Markov bottle. The signal of the CLZ-A pressure-strain type sensor was transformed into digital signal by HX711 24 bit A/D transfer, and then the digital signal was inputted into Arduino UNO interface. FDR soil moisture sensors could give digital signal, which was inputted into Ardunio mega 2560 interface. The 28BYJ-48 byte rotor was controlled by the ULN2003 byte rotor driver, and the 42BYGH 47 byte rotor and 57BYG250B byte rotor were controlled by the TB6600 byte rotor driver. Both ULN2003 byte rotor driver and TB6600 byte rotor driver received control signal that came from Arduino mega 2560 single-chip. The area of FDR soil water moisture sensor was on the surface of a cylinder. When the FDR soil moisture was driven into the soil column in the horizontal direction, there was a one-to-one correspondence between response of FDR soil moisture sensor and water content of the soil column. For this reason, with the infiltration continued, the wetting front moved downward and the detection device obtained the change of soil moisture. According to the 4 basic assumptions of Green-Ampt, when the wetting front had arrived the highest position of area that soil moisture sensor detected, the value of sensor would become bigger until the wetting front had reached the lowest position. When the value change was less than or equaled to 5%, the sensor position adjusting device would adjust the FDR soil moisture sensor position and the next position would be detected. Main interface of upper desk had been designed using the LabVIEW software. To evaluate the device's measuring accuracy, the FDR soil moisture sensor had detected a series of soil samples with different bulk densities (1.15, 1.20, and 1.25 g/cm3). Each soil sample was tested 3 times and the infiltration water head was 10 mm. The response of FDR soil moisture sensor was measured by the device. The results showed that the sensor position adjusting device and soil sample container all could return to the initial position successfully. The success rate of the sensor returning to the initial position and in or out of the soil column was 100%, indicating the reliability of the device for automatic detection. Compared with the soil moisture determined by the oven-drying method, the sensor measurement results had the maximum relative error of -4.4%, suggesting the reliability of soil moisture detection by the sensor. Compared with the wetting front obtained by the labor method, the maximum and the minimum relative distance measurement error of the wetting front position were -12.9% and -4.2%. The maximum relative error for the artificially measured and automatically detected cumulative infiltration was only 2.27% and the maximum root mean square error was 0.65 g. Those results demonstrated that the device developed could be used as the reliable soil column infiltration automatic test platform.