全天候環境下使用前景能量圖分析法進行閘道車輛偵測

Entrance-Gate Vehicle Detection based on Foreground Energy-Map Analysis under Various Lighting and Weather Conditions

周言霖(YAN-LIN CHOU)

指導教授 ： 林道通

國立臺北大學/電機資訊學院/資訊工程學系/碩士(2016年)

摘要

許多汽車服務廠或修車廠等有入口閘道的場所常會裝設感應線圈來偵測是否有車輛進出，感應線圈有分為單感應線圈和雙感應線圈，雙感應線圈可以判斷逆向事件，但是單感應線圈時常會產生誤判，例如行人或是機車經過產生感應、車輛在閘道的逆向與倒車行為、車輛停留在閘道上等情況。近年來閘道附近都會裝設攝影機設備，因此本研究目的在於利用攝影機獲取的影像來進行影像處理，偵測攝影視野中所有車輛。我們在汽車保養廠建立一個健全的車輛偵測系統，這個系統是安裝在戶外環境的所以需要適應各種環境因素，我們利用可適應性背景分割法來提取出前景影像，分析前景特定條件來取得車輛前景。並利用光流法來判斷依循正確方向進入汽車保養廠閘道的車輛，使用中心點計算來排除暫時逗留在閘道的車輛問題。因為車輛偵測系統需要取得含有車牌影像的快照，我們使用前景能量圖分析的輔助在兩張快照中取得正確的快照。最後我們選擇了一個真實的汽車保養廠進行20天的系統可靠度驗證，我們在汽車保養廠的營業時間內獲得了不錯的測試結果，避免了大部分的誤報。我們所以系統目的完全取代感應線圈降低硬體成本並有效增加服務廠的工作效率。本研究方法與步驟主要包括(1)背景分割;(2)光流法;(3)中心點計算;(4)前景能量圖分析。

關鍵字

車輛計數 ； 前景能量圖分析 ； 背景分割 ； 光流法 ； 閘道車輛偵測

並列摘要

Many automobile service shops are often equipped with a sensing coil to detect when a vehicle enters. Sensing coil can be classified as single sensing coil and duel sensing coil. The duel sensing coil can detect the wrong way driving. However, the single sensing coil often produces false alarms. For example, pedestrians and motorcycles might cross the sensing coil, or vehicles might reverse over or stop on the gate. To address these problems, camera equipment is utilized and installed near the entrance gate to detect the incoming vehicles. Hence, the purpose of this study is using the camera equipment to carry out image processing and detect vehicle in the field of view of the camera. The vehicle detection system is implemented in the automobile service shop. This system is implemented for outdoor environment where it needs to adapt to different environmental factors. The background subtraction method is utilized to extract foreground information of moving object from the camera field of view. This foreground image will be identified as vehicle object, if it meets the pre-defined requirements. The optical flow is also applied to determine the direct of the incoming vehicle. We calculate the central point of vehicle to avoid possible error from a vehicle staying too long in the gate way. Because the vehicle detection system needs to take a snapshot on the license plate, the foreground energy-map analysis is purposed to improve the reliability when we choose the correct snapshot. Finally, a 20 days testing is conducted in a real automobile service shop. The results are very positive, most of common errors are avoid. The camera system is designed to replace the sensing coil to reduce the hardware costs and increase the efficiency of the service. The system consists of four components: (1) Background Segmentation, (2) Center Allocation, (3) Optical Flow, and (4) Foreground Energy-Map Analysis.

並列關鍵字

Foreground Energy-Map Analysis ； Background Subtraction ； Optical Flow Method ； Vehicle Counting ； Entrance-Gate Vehicle Detection

參考文獻

[1] Kar, A., Deb, K.: Moving cast shadow detection and removal from video based on hsv color space. In: Proc. Int. Conf. Electrical Engineering and Information Communication Technology (ICEEICT), IEEE (2015) 1–6

[2] Yi, S., Li, H., Wang, X.: Understanding pedestrian behaviors from stationary crowd groups. In: Proc. Computer Vision and Pattern Recognition(CVPR), IEEE (2015) 3488–3496

[4] Zhang, Y., Kiselewich, S.J., Bauson, W.A., Hammoud, R.: Robust moving object detection at distance in the visible spectrum and beyond using a moving camera. In: Proc. Conf. Computer Vision and Pattern Recognition Workshop (CVPRW’06), IEEE (2006) 131

[5] Wang, J., Bebis, G., Miller, R.: Robust video-based surveillance by integrating target detection with tracking. In: Proc. Conf. Computer Vision and Pattern Recognition Workshop (CVPRW’06), IEEE (2006) 137

[6] Sheng, H., Li, C., Wei, Q., Xiong, Z.: Real-time detection of abnormal vehicle events with multi-feature over highway surveillance video. In: Proc. Int. Conf. Intelligent Transportation Systems, IEEE (2009) 550–556