應用機器視覺於化學實驗室藥劑偵測之研究

The Study of Detecting Medicament in Chemical Labs using Machine Vision

王暐晨(Wei-Chen Wang)

指導教授 ： 吳明川

國立臺北科技大學/機電學院/機電整合研究所/碩士(2010年)

https://doi.org/10.6841/NTUT.2010.00216

摘要

現今大學內普遍設有化學實驗室，並備有許多不同種類與性質的化學藥劑，其中有許多對人類與環境具有危害性質，而需要進行管控的化學藥劑。本研究目的基於聯合國主導推行的GHS(Globally Harmonized System of Classification and Labeling of Chemicals)化學品分類與標示全球調和系統，建置一套主動式的化學藥劑偵測系統，利用PTZ IP攝影機監控化學實驗室特定區域並判斷是否具有危險的化學藥劑存在。
本研究中利用數位影像處理技術分離物件與Harr-like特徵進行Adaboost演算法分類訓練偵測化學藥瓶位置，並辨識標籤上9種聯合國所定義之GHS危害圖示，其偵測方法利用輪廓比對技術進行危害圖示偵測。經由實驗的結果可知，化學藥瓶偵測系統與GHS危害圖示偵測系統其偵測時間分別為73ms與42ms，足以應付影像即時監控，進而達成化學實驗室藥劑管理的目標。

關鍵字

化學藥劑 ； Harr-like特徵 ； Adaboost演算法 ； PTZ IP攝影機

並列摘要

Since, there are many kinds of chemical medicaments in the college laboratories. But, some of them which own noxious might damage environment and have to be under control.
This study it based on the globally harmonized system of classification and labeling of chemicals, we build up an active monitoring system which can detect chemical medicaments, and monitoring chemical labs to determine whether the specific region has dangerous medicament.
In this paper, we use digital image processing to separate objects. Then use Harr-likes features with Adaboost algorithm to find chemical bottles location, and detecting nine kinds of hazard pictograms definition in GHS which detecting method is using contour matching techniques. By the result of the experiment, the chemical bottle and the hazard pictogram detecting system take about 73 milliseconds and 42 milliseconds in each frame respectively. It shows that our algorithm able to monitor in real-time, and to reach the goal of chemical laboratory surveillance and management.

並列關鍵字

Chemical Medicament ； Harr-Like Features ； Adaboost Algorithm ； PTZ IP Camera

參考文獻

[3]T. Kawanishi, T. Kurozumi, k. Kashino and S. Takagi, “A fast template matching algorithm with adaptive kipping using inner subtemplates distances,” IEEE Computer Society, vol.3, 2004, pp.654-657.

[4]S. Kaneko, Y. Satoh and S. Igarashi, “Using selective correlation coefficient for robust image registration,” The Journal of The Pattern Recognition society, vol.36, 2003, pp.1165-1173.

[5]D. H. Ballard, “Generalizing the hough transform to detect arbitrary shapes”, Pattern Recognition, vol.13, no. 2, pp.111-122.

[6]V. Ferrari, L. Fevrier, F. Jurie and C. Schmid, “Groups of adjacent contour segments for object detection,” IEEE Computer Society, vol.30, no.1, 2008, pp.36-51.

[8]R. Smith and B. Pham, “A robust object category detection system using deformable shapes,” Machine Vision and Applications, vol.20, 2008, pp.119-130.