倒車輔助攝影機之移動障礙物偵測

Moving Object Detection Using Rear Parking Assist Camera

丁淯斯(Yu-Szu Ting)

指導教授 ： 陳柏全

國立臺北科技大學/機電學院/車輛工程系所/碩士(2008年)

摘要

當車子在路邊停車或倒車入庫時，若有移動障礙物出現在停車路徑中，可利用倒車攝影機拍攝之畫面，將移動障礙物經過辨識後利用方框標示出來。為了實現移動障礙物的偵測解析，本文首先利用偏航角速度資訊計算出先前畫面與目前畫面之運動向量，進行目前畫面之運動補償(motion compensation)，以減少因車輛轉彎造成兩個畫面背景差異性過大的問題，接著用目前畫面減去先前畫面，將前景與背景分離出來，以辨識出移動障礙物，並搭配超音波雷達偵測後方障礙物距離。為了符合高速即時運算的需求，本文利用DSP晶片實踐所提出的策略，並且以實車實驗驗證在路邊停車及倒車入庫時，模擬行人來回在車後方行走、行人來回在車後方回來跑動、行人在後方丟擲籃球後從車後方跑開三種情況。初步結果發現若有進行運動補償，可降低背景差異性太大所造成的誤判；縮小障礙物面積判斷門檻值可提升辨識率，相對也會提高誤判率。

關鍵字

移動障礙物偵測 ； 運動向量 ； 行人偵測 ； 倒車輔助系統

並列摘要

When moving objects appear in the parking trajectory during parallel parking or reverse parking, rear parking assist camera can be used to obtain the image for parking assist system to identify them with square boxes. In order to accomplish the moving object detection, the yaw rate information is used to calculate the motion vector between the previous and current frames. Motion compensation is then implemented to reduce the background difference between these two frames. The foreground is separated from the background by subtracting the previous frame from the current frame. The ultrasonic sensor is also used to detect the distance between the vehicle and rear obstacles. In this study, a digital signal processor (DSP) is utilized to establish a real-time system to implement the proposed algorithm. Two scenarios, parallel parking and reverse parking, are used for the experiment. Three cases, walking pedestrian, running pedestrian, and pedestrian running after throwing a basketball, are verified for each scenario. Preliminary results show that the false detection rate can be reduced greatly using motion compensation. The cognition rate can be increased by reducing the threshold value of the object area. However, the false detection rate will be increased at the same time.

並列關鍵字

moving object detection ； motion vector ； pedestrian detection ； parking assist system

參考文獻

[3] C. Curio, J. Edelbrunner, T. Kalinke, C. Tzomakas and W. V. Seelen, "Walking Pedestrian Recognition," IEEE Transactions on Intelligent Transportation Systems, vol.1, no.3, 2000, pp. 155-163.

[4] 白家榮，十字路口行人之偵測及追蹤，碩士論文，國立臺灣師範大學資訊教育研究所，台北，2001。

[5] A. Broggi, M. Bertozzi, A. Fascioli and M. Sechi, "Shape-based pedestrian detection," Intelligent Vehicles Symposium, vol. 3-5, 2000, pp.215-220.

[6] 吳奇峻，在動態攝影環境下根據時間整合之行人偵測，碩士論文，國立清華大學資訊工程學系，新竹，2005。

[7] 林立鳳，快速行人偵測基於AdaBoost和樣板比對法的駕駛輔助系統，碩士論文，國立中央大學資訊工程研究所，桃園，2007。