The invention relates to a non-destructive detection method for the hardness of a pear. The non-destructive detection method comprises the following steps: measuring a plurality of different types of hyperspectral images of the pear by a hyperspectral imaging system at a visible-near-infrared waveband, correcting by black and white versions, converting the spectral response intensity of each point in the hyperspectral image into unified 0-100% reflectivity images, extracting and coloring the brightness of the image by Labview software, separating different brightness areas, then selecting 10 different points on the same-color area for averaging to represent the reflectivity of the whole pear, combining a reflectivity curve of the pear to extract reflectivity information of each waveband, extracting a characteristic waveband by a PCA, measuring an actual value of the hardness of the pear by a national standard method, establishing a regression model by a PLS algorithm, obtaining various regression equations, and according to the regression equations, calculating the hardness of the pear by measuring a spectrogram of the measured pear. The non-destructive detection method can be widely applied to quality detection of the pear, and the detection process is convenient, fast, non-destructive and accurate.