脑启发的遮挡目标识别模型研究综述OA
A Review of Brain-inspired Models for Occluded Object Recognition
遮挡目标识别是计算机视觉领域的核心挑战之一.目标物体因遮挡导致的特征缺失和混淆严重制约了当前目标识别模型的性能.研究表明人脑视觉系统具有出色的遮挡目标识别能力和对遮挡场景的强适应能力,为构建脑启发的遮挡目标识别模型提供了生物学依据与创新方向.系统梳理了脑启发遮挡目标识别模型的研究进展.首先,对当前主流的遮挡目标识别模型进行系统介绍.其次,聚焦脑启发模型的生物学基础,阐释了视觉皮层层级处理等神经机制在视觉推理中的关键作用.再次,评述模拟生物视觉功能机理的遮挡目标识别模型.最后,总结已有的工作,对当前脑启发的遮挡目标识别模型的发展方向进行讨论,指出对生物机制的探索仍是脑启发模型创新研究的难点.
Occluded object recognition was one of the core challenges in the field of computer vision.Fea-ture loss and confusion caused by object occlusion severely limited the performance of current object recog-nition models.Studies showed that the human visual system possessed excellent capabilities in recognizing occluded objects and strong adaptability to occluded scenes,providing a biological basis and innovative di-rection for constructing brain-inspired occluded object recognition models.The research progress of brain-inspired occluded object recognition models was reviewed.Firstly,the current mainstream occluded object recognition models were introduced.Then,focusing on the biological foundations of brain-inspired models,the key roles of neural mechanisms such as hierarchical processing in the visual cortex in visual reasoning were explained.Furthermore,occluded object recognition models that simulated biological visual mecha-nisms was evaluated.Finally,existing works were summarized and the development directions of current brain-inspired occluded object recognition models were discussed.It was pointed out out that the explora-tion of biological mechanisms remained a major challenge in the innovation of brain-inspired models.
张利朋;李瑞琪;王宁;曹亮;孟振江;李翠霞;胡玉霞
郑州大学 电气与信息工程学院 河南省脑科学与脑机接口重点实验室 河南 郑州 450001郑州大学 电气与信息工程学院 河南省脑科学与脑机接口重点实验室 河南 郑州 450001清华大学 自动化系 北京 100084||中国电子科技集团公司第二十七研究所 河南 郑州 450047海装某军代室 河南 郑州 450000郑州大学 电气与信息工程学院 河南省脑科学与脑机接口重点实验室 河南 郑州 450001郑州大学 网络空间安全学院 河南 郑州 450002郑州大学 电气与信息工程学院 河南省脑科学与脑机接口重点实验室 河南 郑州 450001
信息技术与安全科学
遮挡目标识别脑机制视觉系统脑启发模型
occluded object recognitionbrain mechanismvisual systembrain-inspired model
《郑州大学学报(理学版)》 2026 (5)
1-13,13
河南省科技创新2030重大项目(2022ZD0208500)河南省科技计划项目(202102310210,232102210072)
评论