適應性濾波器應用於移動物件偵測之積體電路架構設計

近年來隨著國際恐怖事件頻傳，公共安全領域的監控系統開始受到各個國家的重視，加上犯罪率不斷攀升，但安全維護人力始終有限，於是發展智慧型即時監控系統取代傳統被動式監控系統來減輕人力負擔是目前的趨勢。　　智慧型即時監控系統第一步是要先偵測出移動物件，對移動物件作出分類、追蹤…等等的處理後，進而適時地對維安人員發出警告，這對於預防犯罪方面有很大的幫助。在即時監控系統應用中，成功地偵測出移動物件是最基本但也是最重要的工作，因此本論文以此為研究主題，將適應性濾波器應用於智慧型監控系統中，提出以最小均方誤差演算法應用於移動物件偵測系統的架構，使其可以針對外變的環境中得到具有適應性的低通濾波器係數，並藉此濾波器來降低影像之解析度與雜訊分佈，以提高偵測物件的正確率並同時減少整體系統的運算量；未來監視系統的影像解析度將會越來越高，此時系統的運算時間會增加而可能導致無法達到即時性，為了加快其整體系統的運算速度，我們使用硬體架構設計的方式來實現本系統。整個偵測系統的設計係採用標準元件庫IC設計流程，以晶片自動化的佈局與繞線方式加以實現並完成後模擬，在工作頻率為50 MHz，若輸入影像解析度為QVGA (320x240)，則每秒能夠處理約1291張影像。若提高影像的解析度為Full HD (1920x1080)，則每秒能夠處理的影像約48張左右，仍然可以達到即時性。此外，我們依據不同的環境，將所提出之方法加以實測，並與其它方法之實驗結果相互分析比較。而透過實驗結果可以證明，本論文所提出之將適應性濾波器應用於移動物件偵測的系統，在外變因素較為複雜的環境下，以不同的測試樣本與其它降低影像解析度的方式互相比較，可以發現所提之方法可以得到較好的偵測結果，其偵測移動物件的成功率約有84.72%。

關鍵字

移動物件追蹤；適應性濾波器；最小均方誤差演算法

並列摘要

Intelligent surveillance system is an important issue for the security consideration in recent years. The first step of the intelligent surveillance system is moving objects detection. A successful detection can reduce redundant data and provide helpful information for post-processing such as moving objects tracking and analysis. Therefore, the moving objects detection of an intelligent surveillance system is a basic but essential task. This thesis presents a new approach of moving objects detection by using the technique of least mean square (LMS) algorithm. This method can adapt the coefficients of low pass filter to the different environments. The low pass filter is used to reduce image resolution and noise effects such as Gaussian noise or fake motions. In the future, the high resolution images for surveillance system will be more common, and the computing time of the system will be longer. The system of high resolution images operated by the software platform may not achieve real-time, so we need to design and implement hardware architecture of the system to reduce the computing time for real-time processing. The proposed approach is further implemented by the VLSI architecture. For the implementation we follow the IC design flow of cell-based IC design. When image size is QVGA (320×240)，the system can reach 1291 FPS (Frame per second).In the same way to design the VLSI architecture, the system can reach 48 FPS as image size is Full HD (1920×1080). The proposed approach is compared with other methods, such as the technique of down-sampling, average filter, and SMDWT. The experimental results show that the accuracy rates of our approach are better than others in different environments. The accuracy rate for moving objects detection of the proposed approach is 84.72%.

並列關鍵字

visual surveillance system ； LMS algorithm ； moving objects detection

參考文獻

[3]F.-H. Cheng and Y.-L. Chen, “Real time multiple objects tracking and identification based on discrete wavelet transform,” Pattern Recognition, vol. 39, no. 3, pp. 1126-1139, June 2006.

Addison-Wesley, 1992.

[8]C.-H. Hsia, J.-M. Guo, and J.-S. Chiang, “Improved low complexity algorithm for 2-D integer lifting-based discrete wavelet transform using symmetric mask-based scheme.” IEEE Transactions on Circuits and Systems for Video Technology, vol. 19, no. 8, pp. 1-7, August 2009.

[12]C.-H. Zhan, X.-H. Duan, S.-Y. Xu, Z. Song, and M. Luo, “An improved moving object detection algorithm based on frame difference and edge detection,” International Conference on Image and Graphics, pp. 519-523, August 2007.

[13]B.-X. Xiao, C. Lu, H. Chen, Y.-F. Yu, and R.-B. Chen, “Moving object detection and recognition based on the frame difference algorithm and moment invariant features,” Chinese Control Conference, pp. 578-581, July 2008.

國際替代計量

適應性濾波器應用於移動物件偵測之積體電路架構設計

全文下載

主題瀏覽