區塊式立體匹配結合倒傳遞類神經網路應用於雙眼立體視覺自動車導航

Combination of Block-based Stereo Matching and BPNN Applied to ALV Guidance Using Binocular Stereo Vision

周博倫(Po-Lun Chou)

指導教授 ： 駱榮欽

國立臺北科技大學/電資學院/電腦與通訊研究所/碩士(2012年)

https://doi.org/10.6841/NTUT.2012.00303

摘要

本論文中，我們提出使用區塊式立體匹配結合倒傳遞類神經網路應用於雙眼立體視覺之室外自動導航車。首先我們將左右影像以區塊式比對，經由不同的區塊大小，比對其對應點的周遭資訊，並藉由區域匹配連續的特性，排除不適當的對應點。接著訓練倒傳遞類神經網路將影像轉換為立體資訊。透過倒傳遞類神經網路誤差收斂後所得到的權重值，反應出左右影像對應點與世界座標的關係。最後將對應點輸入至已學習且收斂的倒傳遞類神經網路，對應的特徵點在實際空間中的位置即可獲得。經過實驗測試結果得知此系統可以實際在室外道路上行走，導航策略也可成功的閃避障礙物，以證明所提方法之可行性。

關鍵字

立體視覺 ； 自動車 ； 類神經網路 ； 物體偵測

並列摘要

In this research, a method of 3D environmental reconstruction is proposed and it used the block-based stereo matching and Back-Propagation Neural Network (BPNN) with binocular stereo vision for outdoor Autonomous Land Vehicle (ALV) guidance. In the study, first, we get left and right images from the binocular cameras to train a BPNN to convert both 2D images into 3D information. In addition, an improved point correspondence method based on block matching is also proposed. The inappropriate corresponding points will be excluded from the continuous feature of stereo matching. Finally, the corresponding feature points are inputted to BPNN for learning and the positions in the real world can be obtained by the BPNN. Several experiments show the proposed method is feasible to avoid collision with moving objects while ALV moving on the way.

並列關鍵字

Stereo Vision ； Autonomous Land Vehicle ； Neural networks ； Object Detection

參考文獻

[1] Rostam Affendi Hamzah, Rosman Abd Rahim and Zarina Mohd Noh, “Sum of Absolute Differences Algorithm in Stereo Correspondence Problem for Stereo Matching in Computer Vision Application,” IEEE Int. Conf. on Computer Science and Information Technology, 2010, Vol.1, pp.652-657.

[2] Qingxiong Yang, Liang Wang, Ruigang Yang, Shengnan Wang, Miao Liao and David Nister, “Real-time Global Stereo Matching Using Hierarchical Belief Propagation,” Proc. Int. Conf. The British Machine Vision Conference, 2006.

[3] Bumsub Ham, Dongbo Min and Kwanghoon Sohn, “Cost Aggregation with Anisotropic Diffusion in Feature Space for Hybrid Stereo Matching,” IEEE Int. Conf. on Image Processing, 2011, pp.3365-3368.

[5] Te-Hsiu Sun, “Improving Stereo Matching Quality with Scanline-Based Asynchronous Hopfield Neural Networks,” Journal of the Chinese Institute of Industrial Engineers, Vol. 24, no.1, pp. 50-59, 2007.

[6] Te-Hsiu Sun, “Stereo Matching Using Synchronous Hopfield Neural Network,” Journal of the Chinese Institute of Industrial Engineers, Vol. 26, no.4, pp. 276-288, 2009.