Technical Abstract: Following the previous research to classify intact mangoes infested with oriental fruit fly from the control ones using near infrared (NIR) spectra acquired by a spot-type handheld NIR instrument, an attempt to improve the sensitivity of the system by employing NIR imaging technology was conducted. Hyperspectral data in the wavelength region of 400 nm to 1000 nm of four infested mangoes and four control mangoes were acquired at 0, 24 and 48 hours after infestation. Each mango carried sixteen wells created by sterile needles. Nine spectra (3x3 pixels) around each well were extracted, converted to absorbance, smoothed and reduced by data selection at 7 data point intervals (approximately 5 nm). Each spectrum was considered an individual sample and fed into a Bayesian discriminant analysis. By using only 3 wavelengths selected from spectra measured at 48 hours after infestation, classification results with 0.9% false negative (range of fruit fly larvae: 8-57) and 5.7% false positive could be obtained. Gray-scale images of a 4 x 4 cm area around each well, for each fruit and at each time interval, were developed calculating a discriminant function based on the distance to control (Dg) and the distance to infested (Db). Fruits infested with fruit flies exhibited a clear distinction of infested area compared to control. This indicates the possibility of developing a low-cost high-speed NIR-image sorting machine for this purpose.