臺灣博碩士論文加值系統

隨著城市化進程的加快和人口的增長，交通管理已成為城市管理中至關重要的一環。本論文旨在開發一套高精準的車輛計數系統，以協助交通監控部門實時監控高速公路車流量，了解平日與假日的車流狀況，以分析出高峰期與離峰期，讓交通監控部門透過本論文的方法研擬出最好的交通政策。此外機車誤闖高速公路是一個嚴重的安全問題，本論文提出了針對機車誤闖的快速回報系統，以減少此類事件的發生。本論文的演算法流程分為四個階段：第一階段是預處理，利用YOLOv7進行車輛偵測並重新訓練模型以劃分車輛類型，以及設立感興趣區域；第二階段是時間序列影像堆疊，透過TSI技術提出time-slice-advance演算法以提升偵測速度與準確度；第三階段是車輛計數，使用掃描線搭配有限狀態機確保車輛計數的準確度；最後是機車警示，採用YOLOv7偵測機車，當偵測到機車會回傳訊息，以降低國道事故的發生。
實驗結果證明，本論文所提出的方法透過11個影像數據集顯示在精確計數具有顯著優勢，在以車輛朝攝影機駛來的正向資料集中Precision可以達到99.7%，Recall可以達到98.9%；在以車輛駛離攝影機的反向資料集中Precision可以達到100%，Recall可以達到99.3%，以呈現出本論文所提出的演算法有著高精準的車輛計數。此外本論文與其他演算法進行相比，也獲得了較高的準確率。機車警示可利用YOLOv7偵測準確的偵測到機車，並提出警示，以減少憾事的發生。

With the acceleration of urbanization and population growth, traffic management has become a crucial aspect of urban administration. This paper aims to develop a highly accurate vehicle counting system to assist traffic monitoring departments in real-time monitoring of highway traffic flow, understanding the differences in traffic patterns between weekdays and holidays, and analyzing peak and off-peak periods. This will help the traffic monitoring department formulate the best traffic policies using the methods proposed in this paper. Additionally, motorcycle intrusions on highways pose a serious safety issue. This paper proposes a quick reporting system for motorcycle intrusions to reduce such incidents. The algorithm process in this paper is divided into four stages: the first stage is preprocessing, utilizing YOLOv7 for vehicle detection and retraining the model to classify vehicle types, and establishing regions of interest; the second stage is time-series image stacking, proposing the time-slice-advance algorithm through TSI technology to improve detection speed and accuracy; the third stage is vehicle counting, using scan lines combined with a finite state machine to ensure counting accuracy; and finally, the fourth stage is motorcycle warning, using YOLOv7 to detect motorcycles, sending alerts when motorcycles are detected to reduce highway accidents.
Experimental results demonstrate that the proposed method exhibits significant advantages in accurate counting through 11 image datasets. In forward datasets where vehicles approach the camera, Precision can reach 99.7%, and Recall can reach 98.9%; in reverse datasets where vehicles move away from the camera, Precision can reach 100%, and Recall can reach 99.3%, showcasing the high accuracy of the proposed vehicle counting algorithm. Moreover, this paper's algorithm achieved higher accuracy compared to other algorithms. The motorcycle warning system accurately detects motorcycles using YOLOv7 and issues warnings, thereby reducing the occurrence of accidents.

摘要..........i
Abstract..........ii
誌謝..........iii
目錄..........iv
表目錄..........vi
圖目錄..........vii
第一章 緒論..........1
1.1 研究背景..........1
1.2 研究動機..........2
1.3 研究目的..........3
1.4 論文架構..........3
第二章 文獻探討..........4
2.1 傳統影像處理偵測..........5
2.1.1 使用基於視覺的時空剖面分析進行交通流量估計和車輛類型分類[7]..........5
2.1.2 基於主成分追蹤背景建模的自動車輛計數方法[8]..........7
2.1.3 用於即時交通監控的低複雜度視覺系統[1]..........8
2.1.4 將外觀和運動與低等級表示相結合，用於車輛計數和交通流分析[9]..........10
2.1.5 使用低秩分解背景減法模型在複雜交通場景中進行即時車輛偵測和計數[10]..........11
2.2 深度影像處理偵測..........12
2.2.1 一種基於車輛計數法的SSD演算法[11]..........12
2.2.2 基於深度神經網路的精確車輛計數方法[12]..........14
2.2.3 一種使用 Mask R-CNN 的高效能車輛計數方法[13]..........15
2.2.4 一種基於卷積神經網路的快速車輛統計和交通量估計方法[14]..........17
2.2.5 利用 CNN 獲得的時空資訊進行車輛偵測和分類[15]..........18
2.2.6 基於卷積神經網路的車輛計數[16]..........20
第三章 研究內容與方法..........22
3.1 預處理..........23
3.1.1 YOLOv7物件偵測模型..........23
3.1.2 樣本採集..........24
3.1.3 樣本標記..........25
3.1.4 模型訓練..........25
3.1.5 設置感興趣區域..........26
3.2 時間序列影像堆疊..........28
3.2.1 time-slice演算法..........28
3.2.2 time-slice-advance演算法..........32
3.2.3 time-slice-advance演算法 – 調整Bounding box..........34
3.2.4 time-slice-advance演算法 – 修改疊加位置..........40
3.3 車輛計數..........41
3.4 機車警示..........47
第四章 實驗與結果..........48
4.1 測試資料集..........48
4.2 評估標準..........50
4.2.1 混淆矩陣(Confusion Matrix)..........50
4.2.2 精確率（Precision）..........51
4.2.3 召回率（Recall）..........51
4.2.4 準確率（Accuracy）..........52
4.3 數據測試集與結果分析..........52
4.4 與其他方法比較..........56
4.5 機車警示..........58
第五章 結論與未來展望..........60
參考文獻..........61
Extended Abstract..........63

[1]J. I. Engel, J. Martín and R. Barco, "A Low-Complexity Vision-Based System for Real-Time Traffic Monitoring," in IEEE Transactions on Intelligent Transportation Systems, vol. 18, no. 5, pp. 1279-1288, May 2017.
[2]Janak D. Trivedi, Sarada Devi Mandalapu, Dhara H. Dave, "Vision-based real-time vehicle detection and vehicle speed measurement using morphology and binary logical operation, "in Journal of Industrial Information Integration, Volume 27, 100280, May 2022
[3]A. Tourani and A. Shahbahrami, "Vehicle counting method based on digital image processing algorithms," 2015 2nd International Conference on Pattern Recognition and Image Analysis (IPRIA), Rasht, Iran, pp. 1-6, 2015
[4]Y. Lecun, L. Bottou, Y. Bengio and P. Haffner, "Gradient-based learning applied to document recognition," in Proceedings of the IEEE, vol. 86, no. 11, pp. 2278-2324, Nov. 1998
[5]Krizhevsky, Alex, Ilya Sutskever and Geoffrey E. Hinton. “ImageNet classification with deep convolutional neural networks.” Communications of the ACM , December 2012
[6]J. Redmon, S. Divvala, R. Girshick and A. Farhadi, "You Only Look Once: Unified, Real-Time Object Detection," 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, pp. 779-788, 2016
[7]Mau-Tsuen Yang, Rang-Kai Jhang and Jia-Sheng Hou, " Traffic flow estimation and vehicle-type classification using vision-based spatial–temporal profile analysis," in IET Computer Vision, vol. 7, Issue 5, pp. 394–404, October 2013
[8]J. Quesada and P. Rodriguez, "Automatic vehicle counting method based on principal component pursuit background modeling," 2016 IEEE International Conference on Image Processing (ICIP), Phoenix, AZ, USA, 2016, pp. 3822-3826
[9]Z. Gao et al., "Synergizing Appearance and Motion With Low Rank Representation for Vehicle Counting and Traffic Flow Analysis," in IEEE Transactions on Intelligent Transportation Systems, vol. 19, no. 8, pp. 2675-2685, Aug. 2018
[10]Honghong Yang, Shiru Qu, "Real-time vehicle detection and counting in complex traffic scenes using background subtraction model with low-rank decomposition," IET Intelligent Transport Systems, Volume12, Issue1, pp. 75-85, February 2018
[11]Q. Chen, N. Huang, J. Zhou and Z. Tan, "An SSD Algorithm Based on Vehicle Counting Method," 2018 37th Chinese Control Conference (CCC), Wuhan, China, 2018, pp. 7673-7677
[12]M. A. Abdelwahab, "Accurate Vehicle Counting Approach Based on Deep Neural Networks," 2019 International Conference on Innovative Trends in Computer Engineering (ITCE), Aswan, Egypt, 2019, pp. 1-5
[13]Z. Al-Ariny, M. A. Abdelwahab, M. Fakhry and E. -S. Hasaneen, "An Efficient Vehicle Counting Method Using Mask R-CNN," 2020 International Conference on Innovative Trends in Communication and Computer Engineering (ITCE), Aswan, Egypt, pp. 232-237, 2020
[14]H. Yang, Y. Zhang, Y. Zhang, H. Meng, S. Li and X. Dai, "A Fast Vehicle Counting and Traffic Volume Estimation Method Based on Convolutional Neural Network," in IEEE Access, vol. 9, pp. 150522-150531, 2021
[15]G.S.R. Satyanarayana, Prashant Deshmukh, Santos Kumar Das, "Vehicle detection and classification with spatio-temporal information obtained from CNN, " Displays,Volume 75, ISSN 0141-9382, 2022
[16]J. M. Anil, L. Mathews, R. Renji, R. M. Jose and S. Thomas, "Vehicle Counting based on Convolution Neural Network," 2023 7th International Conference on Intelligent Computing and Control Systems (ICICCS), Madurai, India, 2023
[17]C. -Y. Wang, A. Bochkovskiy and H. -Y. M. Liao, "YOLOv7: Trainable Bag-of-Freebies Sets New State-of-the-Art for Real-Time Object Detectors," 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Vancouver, BC, Canada, pp. 7464-7475, 2023
[18]Azriel Rosenfeld and John L. Pfaltz, "Sequential Operations in Digital Picture Processing," Journal of the ACM (JACM), Volume 13, Issue 4, October 1966
[19]A. Vakili and J. -C. Gregoire, "Impact of Frame Loss Position on Transmitted Video Quality: Models and Improvements," 2011 Fifth FTRA International Conference on Multimedia and Ubiquitous Engineering, Crete, Greece, pp. 113-118, 2011