Aiming at the problem that multi-categories iris recognition usually takes long time, and is easy interference and inaccuracy of recognition, this paper proposes secondary iris recognition algorithm based on ant colony mutation particle swarm optimization. Firstly, we use principal component analysis to reduce iris noise and redundancy; then use Gabor filtering and Hamming distance for the first recognition, divide iris into eliminated category and undetermined category, eliminate irises which have large difference, narrow recognition range. For the undetermined category, we use Haar wavelet and BP neural network for the sec- ond recognition, determine iris category. Connection weights of neural network and parameters of Gabor filtering are adaptively optimized by ant colony mutation particle swarm optimization. The proposed algo- rithm with many algorithms in different iris libraries are compared. The experimental results show that the proposed algorithm has high recognition performance, good stability, robustness and short time consumption.