本文提出了一种复杂条件下基于子空间梯度方向直方图跟踪的方法，通过大量样本的离线训练构建目标的投影子

空间，并用梯度方向直方图在子空间的投影作为新的目标描述特征．为了满足实时性的要求，采用积分直方图方法

提高粒子特征的计算速度；然后结合粒子滤波方法在子空间中计箅粒子与训练样本集之间的相似度，进而估计目标

的运动参数．实验结果表明，该方法能够在光照变化、噪声干扰、模糊、目标姿态和尺度改变，以及部分遮捎等恶劣条

件下实现准确跟踪，比传统的跟踪方法具有更高的跟踪精度和跟踪鲁棒性，能够满足地面侦察任务在多种复杂条件下对感兴趣目标进行准确跟踪的需求．-This paper presents a complex condition gradient orientation histogram based on subspace tracking methods, through a large number of samples offline training build target subspace projection and gradient orientation histogram in subspace projection as a new target descr iptive characteristics. In order to meet the requirements of real-time, using integral histogram method improves the computing speed of the particle characteristics　then combined with particle filter in subspace meter grate particles and the similarity between the training sample set, and then estimate the target motion parameters. Experimental results show that this method can change in the light, noise, blur, target attitude and scale changes, and take along some cover in bad condition for accurate tracking, than the traditional tracking method has higher tracking accuracy and robustness to meet the ground reconnaissance missions in a variety of complex conditions accurately track targets of interest requirements.