In order to accurately estimate and correct the frequency nonlinearity of voltage controlled oscillator, a novel approach based on contrast optimization of range profiles is proposed, which will perform nonlinearity estimation and correction adaptively. A temperature-varying tuning model is established in the proposed method, the nonlinearity is obtained with a given temperature by using this model. The nonlinearity is removed from the beat signal, and the range profiles of targets are derived by pulse compression. According to the contrast of range profile, the real temperature is iteratively estimated. The contrast of range profile will converge toward the maximum, so an optimal estimation of temperature, which is nonlinearity, is achieved. Simulation and experimental results show that the proposed approach with consideration on environmental temperature variation can estimate and correct the nonlinearity without any additional hardware circuit. Compared with the conventional approaches, this method can not only accurately estimate and track the frequency nonlinearity of voltage controlled oscillator in realtime, but also has low hardware complexity and computing cost, which is the significant to solve the problem of voltage controlled oscillator frequency drift.