近年來汽機車的車禍事故是越來越多,汽機車的數量以及飽和率也不斷提高,以及智慧型手機的普遍使用讓大家對於手機的依賴性越來越高,不管任何時候都要開著螢幕、關注著手機的訊息,因此只要發生一點小事故,後果真是不堪設想。 因此本研究為了解決這種情況,利用深度學習的影像辨識研究出了一個可以偵測這種狀況發生的姿態檢測,那就是利用深度學習中YoloV2(You only look once, Yolo)演算法加上卷積神經網路(Convolutional Neural Network, CNN)紀錄汽車司機的行為,並正確找出司機的違規行為以嚴格規範,改善司機不正常的行為,以降低車禍率為目標。
In recent years, there are more and more automobile accidents, and the number and saturation rate of automobiles and motorcycles are constantly increasing. The widespread use of smart phones has made people more and more dependent on mobile phones. They must be on at all times. Screen, watching messages on your phone, so as long as a small accident occurs, the consequences are really unimaginable. Therefore, in order to solve this situation, this study used deep learning image recognition to develop a gesture detection that can detect this situation, that is, using the YoloV2 (You only look once, Yolo) algorithm in deep learning and plus The Convolutional Neural Network (CNN) records the behaviors of car drivers and correctly finds drivers' violations to strictly regulate them, improve their abnormal behaviors, and reduce the accident rate.