基於CENTRIST特徵實現行人偵測及距離估計

Pedestrian Detection and Range Estimation Based on CENTRIST Descriptor and Implementation

彭冠銓(Kuan-Chuan Peng)

指導教授 ： 傅楸善

國立臺灣大學/電機資訊學院/資訊工程學研究所/碩士(2012年)

https://doi.org/10.6342/NTU.2012.00159

摘要

本論文提出一個基於CENTRIST描述符號的行人偵測和距離估計方法。在相關研究中（如C4行人偵測方法[7]），CENTRIST描述符號只採用固定比例下的特徵而忽略了更大比例下或是全域的特徵。在本論文中，我們採用了四種不同大小的超級區塊設定以克服此問題。我們利用Daimler行人偵測標準檔案[2]來訓練支持向量機的模型並評估其表現。實驗結果顯示我們提出的方法較C4行人偵測方法在錯誤警告方面表現為佳並有可相比的準確度。我們提出的方法除了容易實現外，也很方便用硬體加速。若拍攝場景的立體結構或深度資訊能夠提供，用我們所提出的方法最後結果會更好。

關鍵字

行人偵測 ； 距離估計 ； CENTRIST描述符號

並列摘要

A method of pedestrian detection and range estimation based on CENTRIST descriptor is proposed in this thesis. In related work such as C4 human detection method [7], the CENTRIST descriptor uses only features in a fixed scale, which may omit the features of larger or global structures. In our work, four different settings of the sizes of the superblocks are adopted to overcome such problem. We use Daimler Pedestrian Detection Benchmark [2] to train SVM models and to evaluate the performance. Our experimental results show that our proposed method outperforms C4 human detection method in false alarms with comparable accuracy. Our proposed method is not only easy to implement but also friendly for hardware acceleration and the performance can be better if geometric or stereo information can be accessed.

並列關鍵字

pedestrian detection ； range estimation ； CENTRIST descriptor

參考文獻

[1] N. Dalal and B. Triggs, “Histograms of Oriented Gradients for Human Detection,” Proceedings of IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR), San Diego, CA, vol. 1, pp. 886–893, 2005.

[2] M. Enzweiler and D. M. Gavrila, “Monocular Pedestrian Detection: Survey and Experiments,” IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), vol. 31, no. 12, pp. 2179–2195, 2009.

[3] H. Hirschmüller and D. Scharstein, “Evaluation of Stereo Matching Costs on Images with Radiometric Differences,” IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), vol. 31, no. 9, pp. 1582–1599, 2009.

[5] X. Wang, T. X. Han, and S. Yan, “An HOG-LBP Human Detector with Partial Occlusion Handling,” Proceedings of IEEE International Conference on Computer Vision (ICCV), Kyoto, Japan, pp. 32–39, 2009.

[6] Wikipedia, “Support Vector Machine,” http://en.wikipedia.org/wiki/Support_vector_machine, 2011.