Analyzing the complex relationships between surface environmental factors (slope, aspect, precipitation, and land surface temperature) and soil moisture in mountainous terrain is critical for advancing high-precision optical and radar satellite-based soil moisture inversion. A Convolutional Neural Network for Mountain Environment Based on Representation Learning (CNN MoE) is proposed. Besides, this study utilizes domestic GF-1 remote sensing data and other datasets to investigate the correlations between soil moisture and topographic factors in Chongqing"s central urban area. High-precision soil moisture data derived from inversion, combined with mountain environmental characteristics, were systematically analyzed using grey correlation analysis, ridge regression, and the Kaiser-Meyer-Olkin (KMO) test to evaluate the relationships with elevation, slope, land surface temperature, precipitation, aspect, and distance to rivers. Results demonstrated strong correlations between soil moisture and elevation (0.7) and slope (0.69) under Chongqing"s mountainous conditions, followed by aspect (0.24), precipitation (0.22), and distance to rivers (0.31), with significant spatial heterogeneity in these relationships across regions. The findings provide a scientific foundation for land management and urban planning in mountainous environments with complex terrain characteristics.