發明
中華民國
110139192
I 772208
基於毫米波雷達的人數偵測方法
元智大學
2022/07/21
毫米波雷達會產生三維點雲數據,我們可以利用點雲來得知場景內人數的多寡以及位置。在本文中使用聚類演算法來處理點雲進而計算出場景內的人數,但由於傳統的聚類演算法對於點雲的效果不佳,且點雲易受移動物體的幅度而產生數量不同的數據,導致在使用傳統聚類演算法時,沒有辦法決定出掃描半徑以及最小包含點這兩個參數。故本文改進了聚類演算法中的參數,讓偵測出來的點雲更接近真實的狀態。但由於每幀點雲的數量差距很大,在點雲稀少的情況下聚類演算法無法辨識出範圍內的物體個數,所以採用了累計幀數的方法,來統計出適合的聚類演算法的參數。 而在以往的研究中,大多數關於人數偵測系統都對於識別場景中人的身份沒有過多的說明,而本文中則是利用匈牙利演算法在偵測人數的同時,分別給予身份,而後再利用卡爾曼濾波器進行軌跡的追蹤。 Millimeter wave radar will generate three-dimensional point cloud data. We can use the point cloud to know the number and location of the people in the scene. In this article, a clustering algorithm is used to process the point cloud and then calculate the number of people in the scene, but because the traditional clustering algorithm is not good for the point cloud, and the point cloud is susceptible to the amplitude of moving objects, the number of different Data, when using traditional clustering algorithms, there is no way to determine the two parameters of scan radius and minimum inclusion point. Therefore, this paper improves the parameters in the clustering algorithm to make the detected point cloud closer to the real state. However, due to the large gap in the number of point clouds in each frame, the clustering algorithm cannot identify the number of objects in the range when the point clouds are scarce, so the method of accumulating the number of frames is used to calculate the appropriate clustering algorithm The parameters of the law. In previous studies, most of the people detection systems did not give too much explanation for identifying the identities of the people in the scene. In this article, the Hungarian algorithm is used to detect the number of people while giving the identities separately, and then use them. The Kalman filter performs trajectory tracking.
產學合作組
(03)4638800#2286
版權所有 © 國家科學及技術委員會 National Science and Technology Council All Rights Reserved.
建議使用IE 11或以上版本瀏覽器,最佳瀏覽解析度為1024x768以上|政府網站資料開放宣告
主辦單位:國家科學及技術委員會 執行單位:台灣經濟研究院 網站維護:台灣經濟研究院