In order to detect firmness of hamimelon, a method based on hyper spectral imaging technique is proposed. Effects of different wave bands, different preprocessing methods, different spectral correction methods and different quantitative methods on the predictive accuracy of the model for firmness are compared and analyzed. Experimental results show that in the spectral region between 500 to 820 nm, the partial least square model of the first derivative spectrum processed by standard normal variate gives the best prediction of firmness, with a correction sample correlation coefficient of 0.873, a lower root mean square errors of correction of 4.18 N, a prediction sample correlation coefficient of 0.646 and a root mean square errors of prediction of 6.40 N. The research indicutes that the hyperspectral imaging technique can be used for the nondestructive detection of firmness of hami melon.