可運用於塞車情況下的車輛偵測系統

A Vehicle Detector Approach for Traffic Jam Conditions

目前一般熟知的影像式車輛偵測器，常利用參考背景圖像的方法，切割出移動輪廓，再利用背景更新的觀念，克服因為亮度變化所造成的錯誤偵測。然而，當這些系統在複雜環境下偵測車輛時，將會面臨到嚴峻的考驗，例如，擁擠的車輛或塞車。在塞車發生的情況下，車輛會佔滿整個車道，使得車道背景無法順利被找出或更新，一旦塞車交通狀況解除時，此背景就不再適合用來作為切割移動物體的參考。本篇論文主要的貢獻在於，提出一個在交通壅塞的情況下，不需使用參考背景影像，就能夠有效率偵測車輛的方法。首先利用自動偵測車道所得到之車道線資訊，提出了以車道為基礎的移動物體偵測與切割。在得到車輛的追蹤軌跡之後，計算行車速度、車流量以及車型統計等交通參數。本論文內容包含：自動偵測車道線的方法，以及利用邊際方塊合併（Merged Boundary Box Rule,MBBR）方法偵測車輛，最後透由車輛追蹤法則，達成更新交通參數。一般而言，車輛壅塞的情況大多發生在清晨上班時刻，以及黃昏下班時刻。為了取得更可靠的實驗數據，選取在清晨及黃昏時刻的市區道路以及高速公路作為實驗場景。最後根據實驗結果，本法則無論在高速公路或是市區道路上，都展現了相當高的偵測追蹤率。

Well-known vehicle detectors utilize background extraction methods to segment moving objects, and the background updating concept is applied to overcome the luminance variation which results in error detection. However, these systems will meet a challenge when detecting the vehicles in complex environments, such as heavy traffic or traffic jams. When traffic jams happen, vehicles will cover the road surface so that the background information cannot be smoothly updated. Once the traffic jam is released, the existing background is not suitable for the segmentation of moving objects. The main contribution of this paper is that an efficient vehicle detection approach is proposed to improve the detection accuracy in traffic jam conditions without referring to the background. The automatic land mask decision gives the land information for the detection and segmentation of moving objects. After getting the tracking trajectories, velocities, flows and classifications of vehicles are updated. The proposed approach includes the automatic land detection method, vehicle detection with merged boundary box rule and the tracking algorithm to update traffic information. Generally speaking, traffic jams frequently appear during the morning and evening rush hours. The experimental scenarios are tested in urban roads and the freeways during the morning and evening rush hours to make the experimental results more reliable. Finally, the experimental results show that the proposed methods can improve the detection ratio effectively both in urban roads and freeways.