頁籤選單縮合
題名 | 應用HTN-JPDA類神經網路解決多目標追蹤之問題=Tracking of Multi-Targets Using JPDA Hopfield-Tank Neural Solution |
---|---|
作者 | 蔡樸生; 陳珍源; Tsai, Pu-sheng; Chen, Jen-yang; |
期刊 | 中華技術學院學報 |
出版日期 | 20030400 |
卷期 | 26 2003.04[民92.04] |
頁次 | 頁145-159 |
分類號 | 448.81 |
語文 | chi |
關鍵詞 | 資料關聯; 聯合機率關聯法; 多重假說追蹤法; 霍普菲爾-坦克類神經網路; Data association; Joint probabilistic data association; JPDA; Multiple hypothesis tracking; MHT; Hopfield-tank neural network; HTN; |
中文摘要 | 資料關聯是解決多目標追蹤之核心技術,但大部份的演算法,包括聯合機率關聯法與多重假說追蹤法都過於複雜,不適合實現與達到即時處理的要求。尤其當追蹤目標增加,合理的事件也隨著急速增加,對每一個合理事件分別計算發生的機率,必將造成計算上沉重的負擔。為了有效掌握目標之運動軌跡,本文將採用霍普菲爾-坦克類神經網路,基於HTN網路中每一節點的狀態必朝向能量遞減的方向上變化之特性與強大的平行處理能力,快速解決資料關聯的問題。 |
英文摘要 | The problem of tracking multiple targets in the presence of clutter is addressed. Data association is the most critical part in multi-target tracking; erroneous data associations can result in lost tracks. The joint probabilistic data association (JPDA) and multiple hypotheses tracking (MHT) algorithm have been previously reported to be suitable for this problem. However, the complexity of this algorithm increases rapidly with the number of targets and returns. The computation for probabilities of enormous feasible events becomes v4ry heavy burden. For real-time processing and tracking performance, it seems that parallel structure is a suitable approach. A Hopfield-Tank Network, which consists of many connected processing elements, is capable of parallel computation, and it is suitable for a solution to the data association problem. |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。