An artificial fish swarm algorithm was proposed based on neighborhood orthogonal crossover operator by introducing neighborhood orthogonal crossover operator into the basic artificial fish swarm algorithm. The method of dynamic adjustment of the vision and step of artificial fish to improve the abilities of searching the global and local extremum has been adopted by the algorithm. It made orthogonal crossover operation between the local extremum of artificial fish swarm and the artificial fish swarm itself to produce a few representative optimal individuals which utilized not only its own useful information but also the optimal information of local extremum to improve the fast convergence and the abilities of searching the global extremum. The test of benchmark functions showed that the algorithm improved the performance effectively.