To overcome the shortness of traditional algorithms relying on the point source assumption,a freeform surface optimization method for the extended source was proposed based on the genetic algorithm.The optimal objects involve the radiation distribution function,the offset of origin and the size factor of receiver about the virtual point source.For each individual the reflector can be calculated numerically by solving the partial differential equations established according to the energy conservation law and Snell′s law corresponding to the set of virtual parameters.The individual fitness concerning the uniformity of the object region and the transmission efficiency was obtained according to the simulation results with an actual extended source.The selection,crossover and mutation operations are implemented until the optimization results can meet the requirements.To verify the effectiveness of the proposed method,a uniform lighting system with cone angle of 10° was designed.The contrast experiments are carried out in Trace/Pro with a disc lambertian source with the size of 30 mm in diameter.The light tracing experiments show that the uniformity of the object region increases from 58.3% to 93.5%,and the transmission efficiency increases from 61.8% to 72.9%.Meanwhile,the optical quality has quite good stability at varying distances.The experimental results demonstrate that the method can achieve the adjustment of the spatial distribution of the extended source radiation which has significant guidance for the design of extended source system.