Abstract: Agriculture is a foundational industry that ensures national economy stability and social development．Using of UAV hyperspectral images for precise crop classification and mapping is crucial for agricultural production management and policy-making．According to characteristics of UAV hyperspectral images with both high spectral resolution and high spatial resolution, a crop fine classification method based on wavelet features and neighborhood information was proposed．Firstly, trend and detail features of hyperspectral signals were extracted by multi-level wavelet decomposition, training samples were used to self-validation to filter out optimal feature combination and optimize classifier parameters．Subsequently, neighborhood analysis was introduced to extend feature information of surface features, enhance samples representativeness in areas with significant spatial differences in crops, and obtain optimal wavelet feature set for final classification．Experimental results showed that, compared with conventional original spectral feature classification, classification accuracy using wavelet feature transform increased from 73.90% to 83.68%, an improvement of 9.78 percentage point, combining neighborhood information further increased classification accuracy to 86.29%．This method effectively improved accuracy of crop classification and significantly reduced interference caused by spatial differences such as spectral feature similarities, variations in crop growth, and soil exposure during crop classification process．Effectiveness of wavelet transform combined with neighborhood analysis in hyperspectral image classification was verified．