An interest points detecting algorithm for UAV image is proposed

aiming at problem that UAV reconnaissance image

after using POS parameters to correct spatial-attitude

is easy to be influenced by illumination changes

image blur

image compression and noise in processing. This algorithm firstly calculates offset of barycenter（ Oo B） of local area as the value of interest points response function; then a non-maximum suppression in a 5 × 5 neighborhood is applied to obtain interest points

in the meantime

in order to obtain sub-pixel level localization precision of the points

the barycenter of local area is of the best choice. Through experimental comparison

it shows that this algorithm is more robust to influence of illumination changes

image blur

image compression and noise

so it is more suitable for UAV image interest points detection.