English  |  正體中文  |  简体中文  |  Items with full text/Total items : 62796/95837 (66%)
Visitors : 3639710      Online Users : 363
RC Version 7.0 © Powered By DSPACE, MIT. Enhanced by NTU Library & TKU Library IR team.
Scope Tips:
  • please add "double quotation mark" for query phrases to get precise results
  • please goto advance search for comprehansive author search
  • Adv. Search
    HomeLoginUploadHelpAboutAdminister Goto mobile version
    Please use this identifier to cite or link to this item: https://tkuir.lib.tku.edu.tw/dspace/handle/987654321/52559


    Title: 以色彩為基礎的即時物件辨識系統應用於人形機器人視覺
    Other Titles: A real time color-based object recognition system for humanoid robot vision
    Authors: 張維軒;Chang, Wei-hsuan
    Contributors: 淡江大學電機工程學系碩士班
    余繁;Ye, Fun
    Keywords: 物件辨識;色彩;機器視覺;即時;自適性;解析度;Robot;RoboCup;adaptive resolution method;Object recognition;real time
    Date: 2010
    Issue Date: 2010-09-23 17:55:09 (UTC+8)
    Abstract: 機器人相關研究一直是科技發展的重要議題之一,在眾多種類的機器人當中以自主性人形機器人的發展最為受到關注,目前在機器人學與人工智慧的蓬勃發展之下,機器人能完成的事情不再只是簡單而重複的動作,而是期盼機器人具有獨立思考的能力,身處動態且未知的環境能自行完成高度複雜性的問題,因此機器人之視覺便成為自主性機器人系統最關鍵技術之一。在RoboCup人形機器人足球競賽中,機器視覺常被使用來作目標物辨識、目標物座標判定、動態目標追蹤、目標距離估測、空間定位、路徑規劃、動態障礙物規避以及多個機器足球員間合作進攻防守的智慧型控制策略,因此一個具有即時性與高辨識準確率視覺系統是很被需要的。
    找出畫面中所需要的物件資訊是機器人視覺系統中最基本且最重要的任務,一般來說,物件辨識不外乎是利用其顏色、形狀、輪廓、大小等特徵資訊,將物件從畫面中分割出來,而解析度是影響一張影像中資訊量多寡的關鍵,過大的解析度往往會增加系統運算量的負擔,因此許多降低解析度方法相繼被提出,常見的降低解析度方法有直接次取樣方法、平均濾波次取樣方法、利用二維離散小波轉換降低解析度之方法與對稱式遮罩小波轉換降低解析度之方法。降低解析度不但能加快處理運算的速度,亦能抑制雜訊的干擾,但過度的降低解析度會造成畫面的資訊量不足導致辨識的失敗,因此如何選擇一個適合的解析度大小比起如何降低解析度的方法,是一個更為重要的課題。
    有鑑於此,本論文針對2009年世界盃機器人足球競賽人形機器人足球比賽之規則提出了一套具有自適性解析度調變機制的物件辨識演算法,在比賽中能選擇對整體情勢最為有利的解析度畫面,而利用對稱式遮罩離散小波轉換中的對稱式低低頻遮罩演算法所發展出自適性解析度調變機制,亦同時能抑制雜訊的生成以減少後處理運算量,使得此物件辨識演算法更加的強健。本論文也使用以色彩為基礎的物件分離方法,透過HSV色彩模型以降低對於動態環境光源變化的影響,並加入簡單且快速的物件特徵選取方法,無須透過建立樣板或利用統計等複雜計算,而僅使用其長寬、面積比例等型態特徵便能識別出球場中的關鍵物件。
    而實驗結果證明,本論文提出之物件辨識演算法不但不容易受到亮度變化的影響,且針對球場環境的辨識準確率平均約在93%以上,而平均影格速率能有32fps的水準,不但維持高解析度畫面(320×240)所擁有的高辨識準確率,並提昇其平均影格速率約11fps,改善高解析度所欠缺的即時性,使人形機器人足球員更能得心應手處理球場上瞬息萬變的狀況。
    The research of robots is one of the most important issues in recent years. In the numerous robot researches, the development of autonomous robots is attracted mostly. Accompanying the advancement of robotics and artificial intelligence, the autonomous robots can not only handle simple and monotonous problems, but also have independent thoughts to deal with the complex states in the unpredictable and dynamic environments. All these functions rely on a powerful vision system, and therefore the vision system is one of the most critical techniques for the autonomous robot system. In the RoboCup soccer humanoid league competition, the vision system is used to collect various environment information as the terminal data to finish the functions of object recognition, coordinate building, robot localization, robot tactic, barrier avoiding, etc. Hence a real-time and high recognition accuracy rate vision system is essential.
    Extracting the frame information is the most basic and most important task of the robot vision system. Generally speaking, the object recognition uses object features to extract the object out of the picture frame, and thus color, shape, contour, texture, and sizes of object features are commonly used. The resolution is the key point to affect the frame information. High resolution causes high computing costs, and many low resolution methods were purposed, such as Down-Sampling Method, Average Filter Method, Discrete Wavelet Transform (DWT), and Symmetric Mask-Based DWT Method. These low resolution methods cannot only increase the processing speed, but also reduce the noise interference. However, the reduction of the resolution may reduce the frame information and further fail the recognition. Therefore how to select the most suitable resolution is a more important issue than how to reduce the resolution.
    According to the above-mentioned problems, we propose a new approach, Adaptive Resolution Method (ARM), in the object recognition system for the RoboCup soccer humanoid league rules of the 2009 competition. It can select the most proper resolution for different situations in the competition. The low resolution method is based on the 2-D symmetric mask-based discrete wavelet transform (SMDWT). The SMDWT can reduce the noise interference and make the object recognition system more robustness as well. The HSV color model is used to reduce the influence on the light illumination of the dynamic environment. We also use a simple and fast object feature extraction method to recognize the critical objects by using the features of the shape ratio, area information, etc. The advantage of this method does not need the complexity computing such as the model building, data statistics, etc.
    According to the experimental results, our purposed method is not easily affected by the light illumination. The recognition accuracy rate is more than 93% on average and the average frame rate can reach 32 fps. It does not only maintain the high recognition accuracy rate for the high resolution frames, but also increase the average frame rate for about 11 fps compared to the conventional high resolution approach. It improves the processing time of high resolution to help the humanoid soccer robot to handle the variable situations in the field.
    Appears in Collections:[電機工程學系暨研究所] 學位論文

    Files in This Item:

    File SizeFormat
    index.html0KbHTML391View/Open

    All items in 機構典藏 are protected by copyright, with all rights reserved.


    DSpace Software Copyright © 2002-2004  MIT &  Hewlett-Packard  /   Enhanced by   NTU Library & TKU Library IR teams. Copyright ©   - Feedback