| DC 欄位 |
值 |
語言 |
| DC.contributor | 資訊工程學系 | zh_TW |
| DC.creator | 李資瀚 | zh_TW |
| DC.creator | Zi-Han Li | en_US |
| dc.date.accessioned | 2021-7-19T07:39:07Z | |
| dc.date.available | 2021-7-19T07:39:07Z | |
| dc.date.issued | 2021 | |
| dc.identifier.uri | http://ir.lib.ncu.edu.tw:88/thesis/view_etd.asp?URN=108522043 | |
| dc.contributor.department | 資訊工程學系 | zh_TW |
| DC.description | 國立中央大學 | zh_TW |
| DC.description | National Central University | en_US |
| dc.description.abstract | RPA (Robotic Process Automation) 智慧軟體流程機器人為工業 4.0 中不可或缺的成員之一。RPA 可以應用於高重複性的人工作業流程，透過 RPA 的幫助可以減少人力的耗損以及出錯的可能性。RPA 軟體的應用除了在辦公室自動化之外，還可以應用在工業製造的環境。
本實驗室所開發的 Robotiive 即是適用於工業製造的 RPA 軟體。Robotiive 會擷取當下電腦螢幕畫面，並透過 OpenCV 中的模板匹配 (Template Matching) 的方式，找到使用者所框選的 GUI 元件，並進行操作。
在實際應用的案例中，使用者為了維持機台端電腦的穩定性，在大多數的情況下不會希望將軟體直接安裝在機台端電腦，而是會透過 KVM (Keyboard Video Mouse Switch)、遠端軟體或是網頁來讓 RPA 軟體來對機台電腦進行操作。在這些案例中，我們發現不同的遠端軟體會因為網路因素對回傳的影像進行壓縮，或是網頁文字中產生的反鋸齒 (Anti-aliasing)，造成 Robotiive 的影像辨識失敗。
以上問題雖然透過降低 OpenCV 中模板匹配的閥值或是重新截圖可以改善，不過當環境因素改變後，又會使得新截取的影像辨識失敗。因此，本論文將會以多影像平均值以及高斯濾波基於 OpenCV 的模板匹配，開發出一套不降低閥值且不需要重新截圖就可以正確辨識出目標影像的方法。 | zh_TW |
| dc.description.abstract | RPA (Robotic Process Automation) is an indispensable member of Industry 4.0.RPA can be applied to a highly repetitive manual process. With the help of RPA, attrition and the possibility of human errors can be reduced and automation is enabled. Commercial RPA tools are mainly applied in office automation domain and they are not generally applicable in manufacturing environment.
Robotiive originated from our lab is the RPA software suitable for industrial manufacturing. Robotiive takes a screenshot from current computer and use template matching method in OpenCV to find the GUI component selected by the user.
In practical manufacturing applications, in order to maintain stability of the machine-side computer, users do not want to install the software directly on the machine-side computer in most cases, but will use KVM (Keyboard Video Mouse Switch), remote software or webpage to allow RPA software to operate the machine and computer. In these cases, we found that different remote software
may compress the returned screen image due to network factors, or activate antialiasing effect in the text of the webpage, causing image recognition failed in Robotiive.
Although the above problems can be improved by lowering the threshold of template matching in OpenCV or re-taking screenshot, environmental change will cause the threshold to unstable and sensitive again. Therefore, this thesis use averaging multiple images and Gaussian filter based on OpenCV template
matching to develop a method that can identify the target image correctly without reducing the threshold or re-taking screenshot. | en_US |
| DC.subject | 多影像平均值 | zh_TW |
| DC.subject | 高斯濾波 | zh_TW |
| DC.subject | 影像辨識 | zh_TW |
| DC.subject | Average of multiple image | en_US |
| DC.subject | Gaussian Blur | en_US |
| DC.subject | Image recognition | en_US |
| DC.title | 利用多影像平均值以及高斯濾波方法提升RPA軟體的影像辨識準確度 | zh_TW |
| dc.language.iso | zh-TW | zh-TW |
| DC.type | 博碩士論文 | zh_TW |
| DC.type | thesis | en_US |
| DC.publisher | National Central University | en_US |