Applying pattern recognition methods are important research trend for spectroscopy signal and image processing, one of our researches is putting forward effective qualitative and quantitative analysis method. According to different detection objects and requirements, a variety of spectral and image analysis methods are proposed: Three-dimensional Fluorescence Spectroscopy Discrimination Method for Fat Oxidation of Dairy Product, Training Dictionary and Sparse Representation method for Diesel Property Detection, Wavelet Projection Pursuit Regression Method for Near-infrared Spectroscopy Calibration, Sparse representation classification for Rapid identification of edible oil and swill-cooked dirty oil, and so on. The applicant in nonparametric fitting calculation research for agricultural products, in order to improve qualitative and quantitative prediction performance, verified the feasibility of proposed method.