Abstract: It is possible that the input signal-to-noise ratio of different radars may be different and their reliable estimation may be impossible in practical multiradar distributed detection scenario. So the most robust fusion rule is 1 out of N, which has severe signal-to-noise ratio loss comparable with the whole detection schemes potential. A new type of quarternary local decision based distributed detection algorithm is presented where the fusion center firstly perform three kinds of k out of N fusion and then fuses the obtained decision. Monte Carlo simulation revealed that the proposed algorithm has better detection performance than 1 out of N based scheme and can work well when the input SNRs change violently. So it is a robust distributed detection algorithm.