Based on the idea of the nonlinear dynamics,especially the Thom's Morphogenesis,a knowledge model,called attracting-bifurcating net,is advanced to solve the problem of knowledge representation and acquisition for visual learning in this paper.By introducing the Piaget's concepts in genetic epistemology,the model possesses seven main functions which include consolidatation,assimilation,reunion,accommodation,segmentation and forgetting, thus having given out a completed method for learning visual knowledge.The results three application systems show that the model arid its learning method are effective and adaptable to solve visual problems.