應用於個人醫療裝置之數字辨識

Recognition of Digits Displayed over Personal Healthcare Devices

陳冠廷(Kuan-Ting Chen)

指導教授 ： 張智星

國立臺灣大學/電機資訊學院/資訊工程學研究所/碩士(2019年)

https://doi.org/10.6342/NTU201902097

摘要

本研究的目標是將物件偵測應用在辨識包含個人生理資訊的醫療裝置影像上，藉此取得使用者的生理訊號資訊，如體溫和心率。模型架構是採用在物件偵測中著名的YOLOv3模型，並將之應用在醫療裝置影像中，同時分類和定位醫療裝置之液晶螢幕上的數字及符號，藉此得知使用者的生理資訊，幫助使用者紀錄以及協助使用者判讀數據結果。本研究分為兩部分，分別針對品牌和數字做實驗，品牌部份經過數據增強後結果大幅提升，整體的準確率可以達到99.94%以上，數字方面則是經過影像前處理、刪除重複框選、數字2和5的再辨識和處理小數點問題這些步驟後，準確率從原本的92.66%增長為99.27%以上。

關鍵字

物件偵測 ； 數字辨識 ； 卷積神經網路 ； YOLOv3模型

並列摘要

This paper applies object detection to images from personal healthcare devices in order to extract personal healthcare and physiological information, such as body temperatures and heart rates. The object detection method is based on YOLOv3, which is a successful model for object detection. The adopted method can simultaneously classify and locate the numbers and symbols on the LCD screen of personal healthcare devices, thereby record the personal physiological information on the devices. This thesis is divided into two parts for recognition, including brands and digits. For brand recognition, the average precision is greatly improved after data augmentation, with an average precision over 99.94%. For digit recognition, after performing image preprocessing, deletion of overlapping bounding boxes, reclassification of number 2 and number 5, and reconfirmation of decimal point problem, the average precision can be boosted from 92.66% to 99.27%.

並列關鍵字

Object detection ； Digits recognition ； Convolutional neural network ； YOLOv3

參考文獻

[1] R. Girshick, J. Donahue, T. Darrell, and J. Malik, "Rich feature hierarchies for accurate object detection and semantic segmentation," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2014, pp. 580-587.

[2] J. R. Uijlings, K. E. Van De Sande, T. Gevers, and A. W. J. I. j. o. c. v. Smeulders, "Selective search for object recognition," vol. 104, no. 2, pp. 154-171, 2013.

[3] R. Girshick, "Fast r-cnn," in Proceedings of the IEEE international conference on computer vision, 2015, pp. 1440-1448.

[4] S. Ren, K. He, R. Girshick, and J. Sun, "Faster r-cnn: Towards real-time object detection with region proposal networks," in Advances in neural information processing systems, 2015, pp. 91-99.

[5] W. Liu et al., "Ssd: Single shot multibox detector," in European conference on computer vision, 2016, pp. 21-37: Springer.