Based on mobile robot self-organizing network (Ad-hoc), the dynamic multichannel binocular visual image modeling is studied. By employing three grouped mobile robots with binocular visual system, the true visual images of the scene will be obtained. Then, a three-dimensional model is built by synthesizing the returned images. This paper is aimed at developing a fast three-dimensional modeling algorithm for multichannel binocular visual three-dimensional images, which can improve the accuracy of the dynamic three-dimensional object modeling, reduce the data size and accelerate the modeling speed so as to obtain high speed and accuracy for information obtaining.