基于类脑模型与深度神经网络的目标检测与跟踪技术研究

Object Detection and Tracking Algorithms Based on Brain-Inspired Models and Deep Neural Networks

目标检测与跟踪技术广泛应用于交通、医疗、安保和航天等领域.目前,目标检测与跟踪技术面临目标微弱、背景复杂、目标被遮挡等挑战.同时,随着脑科学研究的不断深入,人们对人脑视觉系统的理解逐渐透彻,利用类脑计算解决复杂背景下高精度目标检测与跟踪问题成为相关领域的重要研究方向.本文结合神经工程导向的类脑模型和计算机工程导向的深度神经网络(Deep Neural Networks,DNNs),提出多种基于类脑模型与深度神经网络的目标检测与跟踪算法,包括:基于演算侧抑制的目标检测算法,基于结构-对比度(Structure-Contrast,SC)视觉注意模型的弱小目标检测算法和基于记忆机制与分层卷积特征的目标跟踪算法.实验结果表明,将类脑模型和深度神经网络应用于目标检测和跟踪领域,有利于实现复杂条件下的高精度目标检测和鲁棒性目标跟踪.

Target detection and tracking technology have been widely used in the fields of transportation,medical,safety and military affairs,etc. However,there still exist some challenges in target detection and tracking,such as dim small target,complex background,target occlusion,and appearance changes,etc. On the other hand,as the most effective bio-intelligence system,Human Visual System has significant advantages in image processing. In this paper,combining neural engineering-oriented brain-like models and computer engineering-oriented DNNs,three target detection and tracking algorithms based on brain-inspired models and DNNs are proposed,including: a moving target detection algorithm based on Algorithmic Lateral Inhibition( ALI) model,a dim target detection algorithm based on Structure-Contrast Visual Attention model,and a target tracking algorithm based on memory mechanism and convolutional feature. The comparison experiments show that applying the brain-inspired models and DNNs to the infrared target detection and tracking is beneficial to achieve accurate target detection and robust tracking under complex conditions.