论文部分内容阅读
以视频速率应用图象识别技术进行识别和跟踪目标是一有广泛兴趣的问题,但过去的尝试只限于很简单的阈值或相关方法。根据新的高速算法和快速数字硬件研制出一种导弹和飞机的识别和跟踪系统。这个系统具有以往实时跟踪系统中没有实现的某种程度的“智能”。它把自适应统计聚类和以投影为基础的分类算法用于实时地识别和跟踪穿过复杂的非平稳背景或前景变化着的目标。组合线性和二次估计的快速估值和予测算法,保证了速度和灵敏度。确定加数,以便提供数据和结果判定的置信度的一种度量。已研究出保持最大跟踪概率的策略。本文着重说明该系统的理论问题和达到实时实现所用的方法。
The use of image recognition at video rates to identify and track targets is a subject of general interest, but past attempts have been limited to very simple thresholds or related methods. According to the new high-speed algorithm and fast digital hardware to develop a missile and aircraft identification and tracking system. This system has some level of “intelligence” that was not realized in real-time tracking systems in the past. It uses adaptive statistical clustering and projection-based classification algorithms to identify and track targets that move through complex non-stationary backgrounds or foreground in real time. The combination of linear and quadratic estimation of the rapid estimation and prediction algorithm to ensure the speed and sensitivity. Additions are determined in order to provide a measure of confidence in the data and outcome decisions. Strategies have been developed to maintain the maximum probability of tracking. This article highlights the theoretical issues of the system and the methods used to achieve it in real time.