基于局部纹理特征的实时目标跟踪系统设计

Design of a real-time target tracking system based on local texture feature

为了有效、稳定、高速地跟踪目标,设计了一套高精度的目标跟踪算法和高性能的目标跟踪硬件系统。为了满足系统高精度跟踪的需要,在深入分析LBP算法优缺点的基础上,提出了一种基于LBP算法与差值匹配算法相结合的改进的纹理特征提取算法;为了满足系统对实时性跟踪的要求,结合算法特点,设计了一套基于高速DSP和FPGA的目标跟踪硬件系统。实验结果表明:改进算法的目标跟踪精度比传统跟踪算法提高很多,而且能够在20 ms内完成目标位置的计算。设计的系统已应用到实际工程中,且能满足目标跟踪系统对实时性、高精度、抗旋转等的要求。

In order to track target effectively,a high precision target tracking algorithm and high performance hardware system are designed.The target feature extraction method and high speed hardware system structure are discussed in detail.In order to meet the requirements of high precision tacking,a proved local texture feature extraction method is presented which combining the proved feature extraction algorithm and different value match methods based on analyzing its advantage and disadvantage thoroughly.In order to meet the requirements of real time processing,a target tracking hardware system based on high performance DSP and FPGA is designed.Experiment results indicate that the precision of this method is better than the traditional methods and the target displacement can be computed in less than 20 ms.The proposed method and hardware system have already be used in actual project which can meet the requirements of the real time processing,high precision and anti-rotation.