隨著車輛數量的增長，交通事故的頻率逐漸上升，這導致車輛損壞的比例也隨之增加。在這樣的背景下，車禍後的理賠流程通常顯得繁瑣而複雜，涉及事故現場圖像的收集、損壞情況的鑑定以及保險公司的評估。整個過程不僅耗時，還需要大量的人力資源，這對於保險公司和車主而言都造成了不便和困擾。
為了解決這一問題，本論文提出了一種基於影像檢測的自動化方法。該方法利用影像角點技術來識別汽車部位的損壞範圍及其大小，從而幫助保險公司迅速判斷損壞部位和範圍。通過自動化的影像分析，該方法能夠顯著縮短理賠時間，提高工作效率。具體而言，保險人員可以依賴此技術快速獲取準確的損壞評估結果，從而減少傳統評估方法中可能出現的人為錯誤。
此外，本研究還探討了如何將該自動化方法應用於實際理賠流程中，以便保險公司和車主能夠更快捷地完成理賠手續。這一創新技術不僅提升了事故評估的準確性，還優化了整體理賠流程，增強了客戶滿意度。總之，本論文為保險行業在事故評估和理賠管理方面提供了一種有效且具有廣泛應用潛力的解決方案，有望推動行業向數字化、自動化轉型邁進。

As the number of vehicles increases, the frequency of traffic accidents has gradually risen, leading to a corresponding increase in vehicle damage. Against this backdrop, the claims process following an accident often appears cumbersome and complex, involving the collection of images from the accident scene, assessment of damage, and evaluation by insurance companies. The entire process is not only time-consuming but also requires significant human resources, causing inconvenience and frustration for both insurance companies and vehicle owners.
To address this issue, this thesis proposes an automated method based on image detection. This method utilizes corner detection techniques to identify the extent and size of damage to vehicle components, thereby assisting insurance companies in quickly determining the damaged areas and their scope. Through automated image analysis, this approach can significantly shorten the claims processing time and improve operational efficiency. Specifically, insurance personnel can rely on this technology to rapidly obtain accurate damage assessment results, thereby reducing potential human errors associated with traditional evaluation methods.
Furthermore, this study explores applying this automated method within actual claims processes so that insurance companies and vehicle owners can complete claims more efficiently. This innovative technology not only enhances the accuracy of accident assessments but also optimizes the overall claims process, increasing customer satisfaction. In summary, this thesis provides an effective solution with broad application potential for the insurance industry in accident assessment and claims management, with the promise of facilitating the industry's transition toward digitalization and automation.

LIST OF CONTENTS

LIST OF CONTENTS	v
LIST OF FIGURES	vii
LIST OF TABLES	ix
CHAPTER 1 INTRODUCTION	1
1.1 Background	1
1.2 Motivation	1
1.3 Target	2
1.4 Research Structure	3
CHAPTER 2 RELATED WORK	4
2.1 Vehicle Damage Detection and Recognition	4
2.2 Harris Corner Detection	5
2.3 Density-Based Spatial Clustering of Applications with Noise (DBSCAN)	7
2.4 Convex Hull	9
CHAPTER 3 METHODOLOGY	12
3.1 System Architecture	12
3.2 Research Process	12
3.3 Data Collection	14
3.4 Data Processing	15
3.5 Car Component Recognition Model	16
3.6 Identifying Damaged Component Models	17
3.6.1 Image Preprocessing	18
3.6.2 Damaged Area Detection	21
3.6.2.1 Detection of Corner Points in Damaged Areas	21
3.6.2.2 Elimination of Similar Corner Points	24
3.6.2.3 Clustering of Corner Points	26
3.6.2.4 Noise Filtering	28
3.7 Calculation of Damage	33
3.7.1 Damage Area Calculation	33
3.8 Show the Results	35
CHAPTER 4 EXPERIMENT AND RESULTS	36
4.1 Data Collection	36
4.2 Data Processing	38
4.3 Car Component Recognition Model	40
4.3.1 Building the Car Component Model	40
4.3.2 Identifying Car Parts	41
4.4 Identifying Damaged Component Models	42
4.4.1 Image Preprocessing	43
4.4.2 Damage Area Detection	45
4.4.2.1 Elimination of Similar Corner Points	46
4.4.2.2 Clustering of Corner Points	48
4.4.2.3 Noise Filtering	49
4.5 Calculation of Damage	52
4.5.1 Damage Area Calculation	52
4.6 Show the Result	53
CHAPTER 5 CONCLUSION	54
5.1 Conclusion	54
5.2 Future Directions	54
REFERENCES	56

LIST OF FIGURES
Fig. 1 System Architecture.	12
Fig. 2 Research Process.	13
Fig. 3 Damage Area Assessment Process.	14
Fig. 4 YOLO Identifies Car Lights.	17
Fig. 5 Left and Right Lights.	18
Fig. 6 Resize the Image to 512x512 pixels.	19
Fig. 7 Background-removed Image.	20
Fig. 8 Grayscale Image.	21
Fig. 9 Harris Corner Detection Schematic Diagram.	22
Fig. 10 Elimination of Similar Corner Diagram.	24
Fig. 11 Eliminate Similar Corner Results.	26
Fig. 12 Clustering Schematic Diagram.	28
Fig. 13 Noise Image.	28
Fig. 14 Outer Boundary of the Car Light (contour area).	29
Fig. 15 Area of each Cluster.	29
Fig. 16 Height and Width of the Image.	34
Fig. 17 Actual Dimensions (length and width) of the Object.	34
Fig. 18 Complete Car Image.	37
Fig. 19 Damaged Car Lights.	38
Fig. 20 Labeling Results of the Left and Right Car Lights.	39
Fig. 21 Data Segmentation.	40
Fig. 22 Results of Car Light Recognition	42
Fig. 23 Cropping Car Lights Image	43
Fig. 24 Resize 512*512 Images.	44
Fig. 25 Background-removed Image.	44
Fig. 26 Grayscale Image.	45
Fig. 27 Harris Corner Detection Results.	46
Fig. 28 Point-to-point Threshold Setting Evaluation Results.	47
Fig. 29 Elimination of Similar Corner Result.	48
Fig. 30 Clustering Image Results.	49
Fig. 31 Cluster Area Calculation Results.	50
Fig. 32 Contour Area Calculation Results.	51
Fig. 33 Percentage Threshold Evaluation Results.	51
Fig. 34 Damage Area Detection Results.	52
Fig. 35 Damage Information.	53

LIST OF TABLES
Table 1. Augmentation Parameters.	39
Table 2. Model Parameter Configuration.	41

REFERENCES
[1]	S. C M, S. Kunjumon and N. R, "Car Damage Identification and Categorization Using Various Transfer Learning Models," 2021 5th International Conference on Trends in Electronics and Informatics (ICOEI), Tirunelveli, India, 2021, pp. 1097-1101, doi: 10.1109/ICOEI51242.2021.9452846.
[2]	X. Wang, W. Li and Z. Wu, "CarDD: A New Dataset for Vision-Based Car Damage Detection," in IEEE Transactions on Intelligent Transportation Systems, vol. 24, no. 7, pp. 7202-7214, July 2023, doi: 10.1109/TITS.2023.3258480.
[3]	Q. Zhang, X. Chang and S. B. Bian, "Vehicle-Damage-Detection Segmentation Algorithm Based on Improved Mask RCNN," in IEEE Access, vol. 8, pp. 6997-7004, 2020, doi: 10.1109/ACCESS.2020.2964055.
[4]	K. Patil, M. Kulkarni, A. Sriraman and S. Karande, "Deep Learning Based Car Damage Classification," 2017 16th IEEE International Conference on Machine Learning and Applications (ICMLA), Cancun, Mexico, 2017, pp. 50-54, doi: 10.1109/ICMLA.2017.0-179. 
[5]	Y. Ma et al., "A System for Automated Vehicle Damage Localization and Severity Estimation Using Deep Learning," in IEEE Transactions on Intelligent Transportation Systems, vol. 25, no. 6, pp. 5627-5639, June 2024, doi: 10.1109/TITS.2023.3334616.
[6]	B. L. Sahu, P. Chandrakar, S. Kumari, C. -M. Chen and M. Amoon, "Blockchain and Machine Learning Integrated Secure Driver Behavior Centric Electric Vehicle Insurance Model," in IEEE Transactions on Intelligent Transportation Systems, vol. 25, no. 11, pp. 19073-19087, Nov. 2024, doi: 10.1109/TITS.2024.3439822.
[7]	D. Mallios, L. Xiaofei, N. McLaughlin, J. M. D. Rincon, C. Galbraith and R. Garland, "Vehicle Damage Severity Estimation for Insurance Operations Using In-The-Wild Mobile Images," in IEEE Access, vol. 11, pp. 78644-78655, 2023, doi: 10.1109/ACCESS.2023.3299223.
[8]	B. Zhang and X. Liu, "Intelligent Pavement Damage Monitoring Research in China," in IEEE Access, vol. 7, pp. 45891-45897, 2019, doi: 10.1109/ACCESS.2019.2905845.
[9]	D. Wang et al., "Combined Use of FCN and Harris Corner Detection for Counting Wheat Ears in Field Conditions," in IEEE Access, vol. 7, pp. 178930-178941, 2019, doi: 10.1109/ACCESS.2019.2958831.
[10]	S. Kye, H. Jeong, D. Cho, Y. Hwang and H. -J. Jung, "Feasibility Study for the Fine Crack Width Estimation of Concrete Structures Based on Fiducial Markers," in IEEE Access, vol. 9, pp. 145238-145247, 2021, doi: 10.1109/ACCESS.2021.3121873.
[11]	Y. Ye, M. Wang, S. Hao and Q. Zhu, "A Novel Keypoint Detector Combining Corners and Blobs for Remote Sensing Image Registration," in IEEE Geoscience and Remote Sensing Letters, vol. 18, no. 3, pp. 451-455, March 2021, doi: 10.1109/LGRS.2020.2980620.
[12]	T. Ouyang, W. Pedrycz, O. F. Reyes-Galaviz and N. J. Pizzi, "Granular Description of Data Structures: A Two-Phase Design," in IEEE Transactions on Cybernetics, vol. 51, no. 4, pp. 1902-1912, April 2021, doi: 10.1109/TCYB.2018.2887115.
[13]	H. Zhang, Z. Duan, N. Zheng, Y. Li, Y. Zeng and W. Shi, "An Efficient Class-Constrained DBSCAN Approach for Large-Scale Point Cloud Clustering," in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 15, pp. 7323-7332, 2022, doi: 10.1109/JSTARS.2022.3201991.
[14]	Z. Zhao, W. Zhou, D. Liang, J. Liu and X. Lee, "Denoising Method Based on Improved DBSCAN for Lidar Point Cloud," in IEEE Access, vol. 12, pp. 137656-137666, 2024, doi: 10.1109/ACCESS.2024.3464611.
[15]	R. Hector, R. Vaidyanathan, G. Sharma and J. L. Trahan, "Optimal Convex Hull Formation on a Grid by Asynchronous Robots With Lights," in IEEE Transactions on Parallel and Distributed Systems, vol. 33, no. 12, pp. 3532-3545, 1 Dec. 2022, doi: 10.1109/TPDS.2022.3158202.
[16]	Q. Sun, Y. Nie, Q. Zhang and G. Li, "Building Coarse to Fine Convex Hulls With Auxiliary Vertices for Palette-Based Image Recoloring," in IEEE Transactions on Visualization and Computer Graphics, vol. 30, no. 8, pp. 5581-5595, Aug. 2024, doi: 10.1109/TVCG.2023.3296386.
[17]	X. Wei, J. Li, B. Long, X. Hu, H. Wu and H. Li, "Method for Vortex Shape Retrieval and Area Calculation Based on Convex Hull Algorithm," in IEEE Access, vol. 9, pp. 2706-2714, 2021, doi: 10.1109/ACCESS.2020.3046651.
[18]	R. V. Chadnov and A. V. Skvortsov, "Convex hull algorithms review," Proceedings. The 8th Russian-Korean International Symposium on Science and Technology, 2004. KORUS 2004., Tomsk, Russia, 2004, pp. 112-115 vol. 2, doi: 10.1109/KORUS.2004.1555560.