Suffering from speckle noise is one of the main obstacles to detect accurate change in synthetic aperture radar (SAR) images, gives rise to synthetic aperture radar image quality degradation. To address this issue, existing deep learning methods typically emphasize spatial information. Inspired by the nonlinear signal processing characteristics, this letter introduces the dendrite neuron model for SAR image change detection task. With a locally nonlinear dendrite structure, dendrite neuron model can adaptively construct the nonlinear effects that these synaptic inputs in the dendrite structure, which is beneficial for describing the contribution of these neighbouring synaptic inputs to the task. Meanwhile, multichannel input signals converge into the cell body by these branches of the dendrite layer, from which different intensities of stimulation on the neuron cell body were aggregated. The parameters involved in this model were searched for the optimal combination using a differential evolution algorithm. Though Simple in structure, visual and quantitative results obtained on three real SAR image data sets have demonstrated the effectiveness and robustness of the dendrite neuron model.