The detection of moisture content in solid particles has important applications in fields such as biological research, coal mine production, and medicine. This article develops a non-contact solid particle moisture sensor (NcSPMS). NcSPMS uses the multi-light source detection method based on the principle of diffuse reflection absorption to detect the moisture content of solid particles with different particle sizes, distributions and colors. Firstly, the stacked LED light source structure (SLED-LSS) was designed to achieve illumination of LED light source for measurement (M-LED) and the LED light source for reference (R-LED) at the same position and exit angle, eliminating the influence of irregular particle size and distribution on diffuse reflected light intensity. Secondly, the stacked LED array light source structure (SLEDA-LSS) was designed to improve the detection light intensity and enable NcSPMS to detect the moisture content of solid particles of different colors. Thirdly, modulated light technology is used for the transmission of moisture content information. Two modulation frequencies were set to modulate the measurement light source and reference light source, and different intensities of square wave currents were determined to drive the M-LED and R-LED to improve detection accuracy. Fourthly, the FFT-KF data processing method was designed to effectively extract the information of diffuse reflection measurement light intensity. Finally, the function expression of moisture content was obtained through cubic fitting, and the detection of moisture content by NcSPMS was achieved. In order to verify the effectiveness of NcSPMS, the NcSPMS control system was designed. The experimental results show that compared with SLED-LSS, SLEDA-LSS improves measurement accuracy; the different current pairs driving M-LED and R-LED affect the measurement accuracy of NcSPMS; NcSPMS can detect the moisture content of white, yellow, green, red, and black sand particles, with the detection range of 0.10–11.00 %. The relative uncertainty obtained by applying NcSPMS to the detection of nickel ore particles is 4.37 %.