To design a near-optimal search path for airborne magnetic anomaly detection against a moving target, an optimum algorithm was proposed based on genetic algorithm with cumulative detection probability as fitness evaluation function.Variable-length, real-number encoding was applied to the chromosome to make it close to real search path.Priori-knowledge of the target was applied to limit the path constrains.The initial populations were generated by an initialization strategy with combination of traditional regular pattern and random search pattern, to ensure the individual diversity and high quality.In the process of genetic operator design, we carried out crossover and mutation strategy based on ellipse constraints to ensure the searcher and target follow physically realizable paths where space and time are continuous.Simulation result shows that:The proposed algorithm has rapid convergence speed and stable performance, and it can improve the overall searching effectiveness greatly.