广义概率数据关联算法作为一种全局次优的数据关联算法，以其新颖的可行性划分规则和较少的计算存储需求引起了人们的广泛关注。为了进一步提高数据关联的精度，从红外凝视系统中目标的特点出发，利用目标方向差和灰度变化因子的概念，得到了量测在目标方向和灰度约束下的互联概率；结合GPDA算法得到的基于距离信息的互联概率，利用加权和的方式即可得到引入方向信息和灰度信息约束后的新的互联概率。该算法不仅充分利用了原GPDA算法的优点，而且通过方向和灰度约束降低了错误关联的概率，使得关联精度得到了进一步的提高。通过该算法与原GPDA和JPDA算法对杂波环境下多目标跟踪性能的对比分析，可以得出该算法在继承了原GPDA算法的基础上，以较小的计算量代价使得跟踪性能得到的很大的改善。

Generalized Probability Data Association (GPDA) algorithm is given more attention in virtue of its novel feasible ruling and less computation burden and less computing memory. In order to improve the precision in data association, this paper introduces target characteristics in the infrared staring system, and acquires the correlation probability of the measurement restricted by target direction and grey scale through utilizing concepts of target direction difference and grey scale change factor. Combined with the correlation probability based on distance information which is obtained from GPDA algorithm, a new correlation probability by weighted sum is obtained. The algorithm not only makes the most of advantages of initial GPDA algorithm, but also reduces incorrect correlation probability through direction and grey scale restriction, which brings higher degree of correlation precision. Comparing with performance of initial GPDA and JPDA algorithms on multi-target tracking in the clutter environment, the algorithm inherited from initial GPDA algorithm significantly improves tracking performance with less calculation.