Of all the animal species in the world, insect species are by far the most numerous. Insect-image recognition systems are conventionally based on color or shape features. However, for species in which the color and shape features are very similar, it is necessary to use additional texture information to ensure a reliable recognition result. In most image-based recognition systems, the texture information is obtained by means of discrete wavelet transformation (DWT). However, the DWT performance is readily affected by noise. Accordingly, the present study proposes an insect image recognition system based on a k-nearest neighbor (k-NN) clustering algorithm, in which the Sobel operator is used to extract the gradient intensity features of the image of interest, and the similarity of the image to known images is calculated using an unmatched-point Hausdorff distance method. The experimental results show that the proposed system has both a short recognition time and a high recognition rate.