In order to solve the problem that the target detection algorithm is difficult to distinguish the mixed pixels and select the threshold value, an adversarial growth (AG) algorithm is proposed according to the similarity of the like pixels. First, the growth tree model is applied to target detection. Then, the AG algorithm is used to improve the growth tree model. Finally, the growth results are obtained under the constraints of the two parameters of omission rate and overlap rate, and the detection results are obtained by further analysis of the growth results. Through the analysis of experimental data, it can be seen that the false alarm rate of AG algorithm is 0.31 percentage points lower than the best result of other four traditional algorithms when the detection probability is 90%. The receiver characteristic curves of the algorithm are all located in the upper left of other algorithms in the four sets of data, which verify the effectiveness of the proposed algorithm, and indicate that the algorithm can better distinguish the mixed pixels, overcome the difficult problem of threshold selection, and improve the efficiency of target detection.