跳到主要內容

臺灣博碩士論文加值系統

(18.97.14.87) 您好!臺灣時間:2025/02/13 05:47
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果 :::

詳目顯示

我願授權國圖
: 
twitterline
研究生:黃俊儒
研究生(外文):HUANG, CHUN-JU
論文名稱:基於LS-R-YOLO神經網路演算法之即時物件偵測系統與晶片設計
論文名稱(外文):Chip and System Design of Real Time Object Detection Based on LS-R-YOLO Network
指導教授:范育成范育成引用關係
指導教授(外文):FAN, YU-CHENG
口試委員:高立人陳彥霖林承鴻
口試委員(外文):Kau, Lih-JenChen, Yen-LinLin, Cheng-Hung
口試日期:2019-06-26
學位類別:碩士
校院名稱:國立臺北科技大學
系所名稱:電子工程系
學門:工程學門
學類:電資工程學類
論文種類:學術論文
論文出版年:2019
畢業學年度:107
語文別:中文
論文頁數:96
中文關鍵詞:物件辨識自駕車神經網路LiDAR
外文關鍵詞:LiDARObject DetectionSelf-DrivingNeural Network
相關次數:
  • 被引用被引用:0
  • 點閱點閱:711
  • 評分評分:
  • 下載下載:0
  • 收藏至我的研究室書目清單書目收藏:0
近年來自駕車發展迅速且逐漸成熟,Google於2014發表首輛無人車雛形,同年Tesla電動車公司為其Model S電動車加入半自動系統,隨後各大公司開始自駕車的研發,自駕車中需要多重感測器,其中LiDAR(Light Detection and Ranging)感測器和左右攝影機為其最主要之感測設備,由於LiDAR感測器可提供360°環境深度資訊,偵測範圍可達120公尺,而左右攝影機可進一步得到前方影像資訊,如此一來便能準確得到自駕車周邊環境模型,方便後段演算法進行路線規劃,因自駕車感測器成本昂貴,近年來學術界大多都利用網路上的自駕車資料庫進行研究,像著名的KITTI Dataset網站便提供各種免費的自駕車感測器資料,其中則以LiDAR點雲圖及左右圖像為主,多數研究都導入類神經網路作為其主要演算法,為求網路架構方便大多只對單一種資料做處理,並無運用到兩種資料特性做演算法上的分配。本論文透過LiDAR點雲圖及圖像資料特性進行演算法上的分配,首先透過LiDAR點雲圖之深度資訊進行初步物件分割,再利用分割完成之點雲圖找到圖像上的目標區域,並將該區域圖像輸入R-YOLO神經網路進行物件偵測,最終輸出圖像辨識和點雲圖辨識結果,本論文演算法可以達到即時物件辨識,且可同時得到LiDAR點雲圖及圖像的辨識結果,此外本論文將R-YOLO之捲積層做成硬體架構,透過標準數位晶片設計流程完成捲積層之晶片。
Recently the progress of self-driving cars become faster and fully developed. Google Company has released the first self-driving car prototype in 2014. In the same year, Tesla improved Model S electrical vehicles with semiautomatic systems. Later on, a lot of companies began to develop self-driving car technology. Self-driving cars need many sensors to detect environment. LiDAR and cameras are the most important sensors of self-driving car. Since LiDAR sensors can provide 360° environmental depth information, the detection range can be up to 120 meters, and cameras can be obtained the front view of car. System can use that information to build the environment model for route planning. Otherwise, the cost of self-driving system is very expensive because the most of researchers use the online dataset to test their algorithm performance. For example, the famous KITTI Dataset website provides a variety of self-driving sensor data for free, including LiDAR point clouds and color images. Most of the researches of self-driving cars only deal with a single type of data, and there are very few applications for two types of data. In this thesis, our algorithm is distributed through point cloud and image characteristics. Firstly, we perform object segmentation through the depth information of the LiDAR point clouds. Secondly, we use the result of object segmentation to find region of interest on the color image and set it as input of R-YOLO neural network for object detection. Finally, our algorithm will simultaneously obtain LiDAR point cloud and image recognition result in real time. Moreover, we will use cell-based design flow to implement high speed convolution chip for better performance of our system.
目錄

摘要 i
ABSTRACT ii
誌謝 iv
目錄 v
表目錄 ix
圖目錄 x
第一章 介紹 1
1.1 簡介 1
1.2 發展與重要性 2
1.3 研究動機 4
1.4 論文章節組成 5
第二章 相關研究與文獻探討 6
2.1 3D光學雷達 (3D LiDAR) 6
2.1.1 機載型光達 (Airborne LiDAR) 6
2.1.2 地面型光達 (Terrestrial LiDAR) 7
2.1.3 移動型光達 (Vehicle LiDAR) 8
2.2 光達點雲圖切割 (LiDAR Segmentation ) 10
2.2.1 地面提取切割 (Segmentation Base on Ground Extraction) 10
2.2.2 二維網格切割 (Segmentation Base on 2D Grid) 11
2.3 類神經網路 (Neural Network) 12
2.3.1 捲積神經網路 (Convolutional Neural Network) 12
2.3.2 區域捲積神經網路 (Region-based CNN) 15
2.3.3 快速區域捲積神經網路 (Faster R-CNN) 16
2.3.4 YOLO神經網路 (You Only Look Once) 17
2.4 物件辨識演算法 19
2.4.1 純影像物件辨識 19
2.4.2 影像搭配點雲圖之物件辨識 20
2.5 結論 21
第三章 研究方法 22
3.1 系統架構 22
3.2 線上資料集 (Dataset) 23
3.2.1 KITTI Dataset 23
3.2.2 Pascal VOC Dataset 25
3.3 光達切割演算法(LiDAR Segmentation) 27
3.3.1 階層分割(Layer Segmentation) 27
3.3.2 階層合併(Layer Combination) 28
3.3.3 地面去除(Ground Extraction) 29
3.4 二維ROI提取(2D Range of Interest Extraction) 30
3.5 神經網路架構(Neural Network Architecture) 31
3.6 結論 33
第四章 硬體架構設計與實現 35
4.1 硬體系統架構 35
4.2 輸入資料之型態 36
4.3 主控制器架構 38
4.4 電路讀取流程介紹 39
4.5 捲積電路架構介紹 42
4.6 資料流處理 44
4.7 結論 45
第五章 實驗方法與結果 46
5.1 實驗方法 46
5.2 實驗結果 47
5.3 實驗結果比較 66
5.3.1 LS-R-YOLO優勢比較 67
5.3.2 相似演算法比較 68
5.3.3 運算速率比較 69
5.3.4 辨識率比較 70
5.4 結論 71
第六章 晶片設計流程 72
6.1 數位電路設計流程 (Cell-Based Design Flow) 72
6.1.1 晶片腳位定義 (Chip Specifications) 73
6.1.2 RTL設計與模擬 74
6.1.3 邏輯合成 (Logical Synthesis) 75
6.1.4 可測試性電路設計 (Design for Testing) 77
6.1.5 自動佈局與繞線 (Auto Place and Route) 78
6.1.6 DRC與LVS驗證 79
6.1.7 Nanosim 80
6.2 晶片實體佈局與規格 81
6.3 數位晶片量測流程 85
6.4 結論 87
第七章 總結與未來展望 88
7.1 總結 88
7.2 未來展望 89
參考文獻 90
附錄A發表論文 97
附錄B量測報告 98



參考文獻

[1] WAYMO. [Online]. Available: https://waymo.com/
[2] NavLab11. [Online]. Available: http://www.cs.cmu.edu/~tyata/Project/NavLab11.html
[3] GOOGLE'S AUTONOMOUS PRIUS. [Online]. Available: https://www.wired.com/2012/03/googles-autonomous-prius-drives-blind-man-to-taco-bell/
[4] Self-Driving "Firefly" Pod Cars. [Online]. Available:
https://www.motor1.com/news/148354/waymo-self-driving-cars-retired/
[5] Society of Automotive Engineers Automation Levels. [Online]. Available:
https://www.sae.org/news/press-room/2018/12/sae-international-releases-updated-visual-chart-for-its-%E2%80%9Clevels-of-driving-automation%E2%80%9D-standard-for-self-driving-vehicles
[6] Airborne Laser. [Online]. Available:
http://www.grounddatasolutions.com/airbornelaser.html
[7] Spatial variability. [Online]. Available: https://www.researchgate.net/figure/Spatial-variability-of-A-gayanus-cover-mapped-using-airborne-LiDAR-showing-the-advancing_fig6_275517307
[8] Terrestrial Lidar Scanning Research. [Online]. Available:
http://sites.bu.edu/lidar/
[9] Terrestrial Lidar Scanning. [Online]. Available: http://www.ollerhead.ca/technology/terrestrial-lidar-scanning/
[10] Velodyne Lidar. [Online]. Available: https://velodynelidar.com/hdl-64e.html
[11] Point Cloud Library with Velodyne LiDAR. [Online]. Available: http://unanancyowen.com/en/pcl-with-velodyne/
[12] Z. Chen, J. Zhang, and D Tao, “Progressive LiDAR Adaptation for Road Detection,” in IEEE/CAA Journal of Automatica Sinica, vol. 6, pp. 693-702, 2019.
[13] L. Caltagirone, M. Bellone, L. Svensson, and M. Wahde. “Lidar-camera fusion for road detection using fully convolutional neural networks,” arXiv preprint arXiv:1809.07941, 2018.
[14] X. Han, J. Lu, C. Zhao, S. You, and H. Li, “Semi-supervised and weakly-supervised road detection based on generative adversarial networks,” in IEEE Signal Process. Lett., vol. 25, pp. 551-555 2018.
[15] B. Douillard, J. Underwood, N. Kuntz, V. Vlaskine, A. Quadros, P. Morton, and A. Frenkel, “On the Segmentation of 3D LIDAR Point Clouds,” in 2011 IEEE International Conference on Robotics and Automation, Shanghai, China, May 9-13, 2011, pp. 2798-2805.
[16] M. Himmelsbach, F. V. Hundelshausen, and H.-J. Wuensche, “Fast segmentation of 3D point clouds for ground vehicles,” in 2010 IEEE Intelligent Vehicles Symposium, San Diego, CA, USA, Jun. 21-24, 2010, pp. 560–565.
[17] M. Himmelsbach and H. J. Wuensche, “Tracking and classification of arbitrary objects with bottom-up/top-down detection,” in 2012 IEEE Intelligent Vehicles Symposium, Spain, Jun. 3-7, 2012, pp. 577–582.
[18] J. Cheng, Z. Xiang, T. Cao, and J. Liu, “Robust vehicle detection using 3D Lidar under complex urban environment,” in 2014 IEEE International Conference on Robotics and Automation, Hong Kong, China, May 2014, pp. 691–696.
[19] M. Himmelsbach, T. Luettel, and H. Wuensche, “Real-time object classification in 3D point clouds using point feature histograms,” in 2009 IEEE/RSJ International Conference on Intelligent Robots and Systems, MO, USA, Oct. 10-15, 2009, pp. 994–1000.
[20] K. Kidono, T. Miyasaka, A. Watanabe, T. Naito, and J. Miura, “Pedestrian recognition using high-definition LIDAR,” in 2011 IEEE Intelligent Vehicles Symposium, Baden-Baden, Germany, Jun. 5-9, 2011, pp. 405–410.
[21] A. Teichman, J. Levinson, and S. Thrun, “Towards 3D object recognition via classification of arbitrary object tracks,” in 2011 IEEE International Conference on Robotics and Automation, Shanghai, China, May 9-13, 2011, pp. 4034–4041.
[22] D. Held, J. Levinson, S. Thrun, and S. Savarese, “Combining 3D shape, color, and motion for robust anytime tracking,” in 2014 Robotics: Science and Systems Conference, Berkeley, CA, USA, Jul. 12-16, 2014.
[23] An intuitive guide to Convolutional Neural Networks. [Online]. Available:
https://www.freecodecamp.org/news/an-intuitive-guide-to-convolutional-neural-networks-260c2de0a050/
[24] Deep Learning learns layers of feature. [Online]. Available: https://github.com/NirViaje/DeepLearningRobotics/blob/master/NeuronTalk.md
[25] Convolutional neural network. [Online]. Available:
https://en.wikipedia.org/wiki/Convolutional_neural_network
[26] Sigmoid function. [Online]. Available:
https://en.wikipedia.org/wiki/Sigmoid_function
[27] Rectifier (neural networks). [Online]. Available:
https://en.wikipedia.org/wiki/Rectifier_(neural_networks)
[28] Regression Loss Functions. [Online]. Available:
https://heartbeat.fritz.ai/5-regression-loss-functions-all-machine-learners-should-know-4fb140e9d4b0
[29] R. Girshick, J. Donahue, T. Darrell, and J. Malik. “Rich feature hierarchies for accurate object detection and semantic segmentation,” in 2014 IEEE Conference on Computer Vision and Pattern Recognition, Columbus, OH, USA, Jun. 23-28, 2014, pp. 580-587.
[30] M. V, V. V.R, and N. A, “A Deep Learning RCNN Approach for Vehicle Recognition in Traffic Surveillance System,” in 2019 International Conference on Communication and Signal Processing, Chennai, India, Apr. 4-6, 2019.
[31] X. Ren, S. Du, and Y. Zheng, “Parallel RCNN: A deep learning method for people detection using RGB-D images,” in 2017 10th International Congress on Image and Signal Processing, Shanghai, China, Oct. 14-16, 2017.
[32] M. Braun, Q. Rao, Y. Wang, and F. Flohr, “Pose-rcnn: Joint object detection and pose estimation using 3d object proposals,” in 2016 IEEE 19th International Conference on Intelligent Transportation Systems, Nov. 1-4, 2016, pp. 1546–1551.
[33] S. Shi, X. Wang, and H. Li, “Pointrcnn: 3d object proposal generation and detection from point cloud,” arXiv preprint arXiv:1812.04244, 2018.
[34] S. Ren, K. He, R. Girshick, and J. Sun, “Faster R-CNN: Towards real-time object detection with region proposal networks,” in IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 39, pp. 1137-1149, 2017.
[35] K. He, X. Zhang, S. Ren, and J. Sun, “Spatial pyramid pooling in deep convolutional networks for visual recognition,” in IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 37, pp. 1904-1916, 2015.
[36] R. Zhang and Y. Yang, “Merging recovery feature network to faster RCNN for low-resolution images detection,” in 2017 IEEE Global Conference on Signal and Information Processing, Montreal, QC, Canada, Nov. 14-16, 2017.
[37] C. Fu, W. Si, Q. Lu, C. Shi, Q. Gao, H. Wang, and C. Wang, “Study of a Detection and Recognition Algorithm for High-Voltage Switch Cabinet Based on Deep Learning with an Improved Faster-RCNN,” in 2018 International Conference on Engineering Simulation and Intelligent Control, Changsha, China, Aug. 10-11, 2018.
[38] J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, “You only look once: Unified, real-time object detection,” arXiv preprint arXiv:1506.02640, 2015.
[39] J. Redmon and A. Farhadi, “Yolo9000: Better, faster, stronger,” arXiv preprint arXiv:1612.08242, 2016.
[40] J. Redmon and A. Farhadi, “Yolov3: An incremental improvement,” arXiv preprint arXiv:1804.02767, 2018.
[41] T.-Y. Lin, P. Dollar, R. Girshick, K. He, B. Hariharan, and S. Belongie, “Feature pyramid networks for object detection,” arXiv preprint arXiv:1612.03144, 2016.
[42] YOLO: Real-Time Object Detection. [Online]. Available:
https://pjreddie.com/darknet/yolo/
[43] F. Yang, W. Choi, and Y. Lin, “Exploit All the Layers: Fast and Accurate CNN Object Detector with Scale Dependent Pooling and Cascaded Rejection Classifiers,” in 2016 IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, NV, USA, Jun. 27-30, 2016.
[44] J. Ren, X. Chen, J. Liu, W. Sun, J. Pang, Q. Yan, Y. -W. Tai, and L. Xu, “Accurate Single Stage Detector Using Recurrent Rolling Convolution,” in 2017 IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, HI, USA, Jul. 21-26, 2017.
[45] M. Liang, B. Yang, Y. Chen, R. Hu, and R. Urtasun, “Multitask multi-sensor fusion for 3d object detection,” in IEEE Conference on Computer Vision and Pattern Recognition, 2019.
[46] X. Du1, M. H, A. Jr, and D. Rus, “Car Detection for Autonomous Vehicle: LIDAR and Vision Fusion Approach Through Deep Learning Framework,” in 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems, Vancouver, BC, Canada, Sept. 24-28 2017.
[47] KITTI Dataset. [Online]. Available: http://www.cvlibs.net/datasets/kitti/
[48] Pascal VOC Dataset. [Online]. Available: http://host.robots.ox.ac.uk/pascal/VOC/
[49] yolov3-tiny.cfg. [Online]. Available:
https://github.com/pjreddie/darknet/blob/master/cfg/yolov3-tiny.cfg
[50] HDL-64E. [Online]. Available: https://velodynelidar.com/hdl-64e.html
[51] Object Detection Evaluation 2012. [Online]. Available:
http://www.cvlibs.net/datasets/kitti/eval_object.php?obj_benchmark=2d
[52] M. Oeljeklaus, F. Hoffmann, and T. Bertram1, “A Fast Multi-Task CNN for Spatial Understanding of Traffic Scenes,” in IEEE 2018 21st International Conference on Intelligent Transportation Systems, Maui, HI, USA, Nov. 4-7, 2018.
[53] M. C. Chang, Z. G. Pan, and J. L. Chen, “Hardware Accelerator for Boosting Convolution Computation in Image Classification Applications,” in Global Conference on Consumer Electronics, Nagoya, Japan, Oct. 24-27, 2017.
[54] H. M. Chang, M. H. Sunwoo, “An Efficient Programmable 2-D Convolver Chip,” in IEEE International Symposium on Circuits and Systems, CA, USA, May 31-June 3, 1998, pp. 429-432.
[55] K. Benkrid, S. Belkacemi, “Design and Implementation of a 2D Convolution Core for Video Applications on FPGAs,” in Third International Workshop on Digital and Computational Video, FL, USA, Nov. 15-15, 2002, pp. 85-92.
[56] Y. -H. Chen, T. Krishna, J. S. Emer, and V. Sze, “Eyeriss: An Energy-Efficient Reconfigurable Accelerator for Deep Convolutional Neural Networks,” in IEEE Journal of Solid-State Circuits, vol. 52, pp. 127-138, 2017.
[57] S. Li, O. Ning, and Z. Wang, “Accelerator Design for Convolutional Neural Network with Vertical Data Streaming,” in 2018 IEEE Asia Pacific Conference on Circuits and Systems, Chengdu, China, Oct. 26-30, 2018.
[58] VGG16 – Convolutional Network for Classification and Detection, [Online]. Available: https://neurohive.io/en/popular-networks/vgg16/


QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top
無相關期刊