本文選題：盆花 + 圖像處理　； 參考：《天津理工大學》2015年碩士論文

【摘要】：隨著傳感技術、計算機技術、人工智能及其它相關學科的迅速發(fā)展,機器人正向具有自組織、自學習、自適應能力的智能化方向發(fā)展。對于視覺機器人而言,能夠實現(xiàn)對周圍環(huán)境物體的準確識別、定位、跟蹤是代表其智能化能力的關鍵指標。機器人的導航方式有多種,其中,視覺導航系統(tǒng)的研究是當今機器人研究的熱點。在機器人行為控制及視覺導航研究中,如何提高機器人在未知環(huán)境中行為的正確性和視覺系統(tǒng)圖像識別的實時性和準確性,是研究的熱點之一。針對這些問題,本文提出了基于DSP-FPGA的高速圖像采集處理系統(tǒng)設計方案并將其運用在溫室盆花裝載中,該方案對采集到的圖像進行邊緣提取、銳化等圖像處理,并根據(jù)FPGA并行處理的特點提出一種改進的中值濾波方法,且進行仿真實驗,仿真結果表明:在FPGA進行圖像處理中采用改進的中值濾波算法,不僅能夠很好的對采集到的圖片進行去噪,而且具有很快的運算速度。機器視覺定位系統(tǒng)解決機器人的目標定位和跟蹤問題,是整個機器人系統(tǒng)的核心和關鍵。本課題采用跟蹤物體目標的特征點的視覺定位算法為基礎,從攝像機的標定、模板匹配、背景建模、前景目標分離到特征點提取、運動估計及卡爾曼濾波,最終實現(xiàn)了視覺機器人定位的過程及其效果。通過對傳統(tǒng)的角點檢測算法進行深入了解與分析,本文在其基礎上提出了改進。引入一種基于改進Harris角點提取的準確跟蹤方法,該方法在傳統(tǒng)Harris特征點檢測基礎上,利用角點附近像素灰度值梯度的變化關系,利用簡單的運算與分析,首先排除一些偽角點與非角點,接下來再對保留的點做進一步處理,得出正確特征點。通過編寫本算法代碼最終實現(xiàn)其檢測效果,并與傳統(tǒng)算法相比較,得出本算法能夠再更短的時間內(nèi)提取出更為精確的角點,為下一步盆花的準確跟蹤打下的基礎,體現(xiàn)了該算法的實用性。
[Abstract]:With the rapid development of sensing technology, computer technology, artificial intelligence and other related disciplines, robot is developing intelligently with self-organizing, self-learning and adaptive ability. For the visual robot, it is the key index to realize the accurate recognition, location and tracking of the surrounding objects. There are many navigation modes of robot, among which, the research of visual navigation system is the hot spot of robot research. In the research of robot behavior control and visual navigation, how to improve the correctness of robot behavior in unknown environment and the real-time and accuracy of visual system image recognition is one of the hot research topics. Aiming at these problems, this paper puts forward the design scheme of high speed image acquisition and processing system based on DSP-FPGA and applies it to the loading of greenhouse potted flowers. According to the characteristics of FPGA parallel processing, an improved median filtering method is proposed and simulated. The simulation results show that the improved median filtering algorithm is used in FPGA image processing. Not only can the collected images be de-noised very well, but also it has a fast computing speed. The machine vision positioning system is the core and key of the whole robot system to solve the problem of target location and tracking. Based on the visual localization algorithm for tracking the feature points of the object, the subject includes camera calibration, template matching, background modeling, extraction of feature points from foreground targets, motion estimation and Kalman filter. Finally, the process and effect of vision robot localization are realized. Based on the deep understanding and analysis of the traditional corner detection algorithm, this paper proposes some improvements. An accurate tracking method based on improved Harris corner extraction is introduced. On the basis of traditional Harris feature point detection, the change relation of pixel grayscale gradient near corner is used, and simple operation and analysis are used. Firstly, some pseudo-corner points and non-corner points are excluded, then the reserved points are further treated and the correct feature points are obtained. Compared with the traditional algorithm, the algorithm can extract more accurate corner points in a shorter time, and lay the foundation for the further accurate tracking of potted flowers. It reflects the practicability of the algorithm.
【學位授予單位】：天津理工大學
【學位級別】：碩士
【學位授予年份】：2015
【分類號】：TP391.41

【參考文獻】

相關期刊論文 前10條

1 陳俊周;李煒;王春瑤;;一種動態(tài)場景下的視頻前景目標分割方法[J];電子科技大學學報;2014年02期

2 陳汗青;萬艷玲;王國剛;;數(shù)字圖像處理技術研究進展[J];工業(yè)控制計算機;2013年01期

3 辜松;楊艷麗;張躍峰;;荷蘭溫室盆花自動化生產(chǎn)裝備系統(tǒng)的發(fā)展現(xiàn)狀[J];農(nóng)業(yè)工程學報;2012年19期

4 李立芳;;淺談數(shù)字圖像處理技術及應用[J];中國科技信息;2012年03期

5 張秀榮;孟和達來;;數(shù)字圖像處理系統(tǒng)的開發(fā)與研究[J];湖南師范大學自然科學學報;2011年06期

6 許允喜;蔣云良;陳方;;基于2D-3D雙目運動估計的立體視覺定位算法[J];光電工程;2011年09期

7 張建中;何永義;李軍;;機器人裝配視覺定位應用研究[J];機電工程;2011年08期

8 俞吉波;孔雪;鄭哲;祝永新;付宇卓;;FPGA實際可用性評估與發(fā)展趨勢分析[J];計算機工程;2011年13期

9 楊明;宋麗華;;改進的快速中值濾波算法在圖像去噪中的應用[J];測繪工程;2011年03期

10 景敏;;數(shù)字圖像處理技術的應用[J];機械管理開發(fā);2011年01期

，

本文編號：1983074