定點式多攝像頭應用於老人跌倒預防系統之可行性研究

全球人口老化問題日趨嚴重，此趨勢將導致老人照護的需求大幅增加，但面臨照護人力資源的短缺及人力成本的增加，不論是在宅安養還是機構安養，都將需要另謀解決辦法。隨著進入物聯網及建築自動化的時代，無人監控、遠端控制、視覺辨識等技術其成本逐年降低，在建築領域的應用越來越普及。此論文希望透過選定適合的感測技術和硬體設備，並針對當前在老人照護資源短缺的狀況下所產生的問題，設計出一套系統來解決當今社會都要面臨的窘境。目前市面上及學術研究上關於智慧老人跌倒防治的措施較多為事件發生後的感測，本研究將著重於跌倒的事前分析及預警，所謂預防勝於治療，事前預警的優點在於可以避免跌倒事件中醫療人力及資源的消耗。研究將以老人容易因物體絆到導致跌倒為解決對象，以攝像頭為感測器，用程式語言編寫視覺辨識之功能，探討是否能用此設備及技術，良好的接收人物在空間中的移動方向、速度及路徑，並判別是否有可能導致跌倒之物體在移動路徑上，在達到一定條件時發出警報，以避免跌倒發生。本研究主要步驟包括程式開發、硬體測試、模擬實驗、實驗校正和比較與多區域佈點設計，研究內容如下： 1.透過文獻回顧建立現有智慧老人防跌倒的技術及應用，整理智慧老人防跌的應用技術及是屬於事前預防還是事後預警，並調查老人跌倒的主因即發生地點。 2.依照以掌握的軟硬體技術進行老人防跌預警的程式設計，產出一套可調整的系統。 3.設計一系列實驗用來測試並校正防跌系統，使用預先錄好的影片讓系統讀取，模擬即時偵測的情境，並記錄分析實驗數據，根據數據進行調整與比較。 4.根據先前實驗結果，進行多區域的統合性偵測實驗，模擬一般住宅常見臥室擺設狀況下，系統的應變能力。

關鍵字

電腦視覺；老人跌倒；高齡社會；跌倒預警；智慧建築；物聯網

並列摘要

The global population aging problem is becoming more and more serious. This trend will lead to a significant increase in the demand for elderly care. However, facing the shortage of human resources for care and the increase in labor costs, whether it is in home security or institutional security, we will need to find another solution. With the entry into the era of Internet of Things and building automation, the costs of technologies such as unmanned monitoring, remote control, and visual identification have been reduced year by year, and their application in the field of construction is becoming more and more popular. This paper hopes to design a system to solve the dilemma faced by today's society by selecting suitable sensing technology and hardware equipment and addressing the problems arising from the current shortage of elderly care resources. At present, the measures for prevention and control of falls of smart elderly people in the market and academic research are mostly after the event. This study will focus on the pre-analysis and early warning of falls. The so-called prevention is better than treatment. The advantage of pre-warning is that you can avoid falls. The consumption of medical manpower and resources in the incident. The study will take the elderly as a result of falling over easily because of objects tripping, and use the camera as a sensor to write a visual recognition function in programming language, and explore whether this equipment and technology can be used to well receive the direction of the character's movement in space , Speed and path, and determine whether there is an object that may cause a fall on the moving path, and an alarm is issued when certain conditions are reached to avoid a fall. The main steps of this research include program development, hardware testing, simulation experiments, experimental correction and comparison, and multi-region layout design. The research contents are as follows: 1. Through literature review, establish existing technologies and applications for preventing the fall of the smart elderly, sort out the application technologies of the falling prevention of the smart elderly and whether they belong to pre-event prevention or post-warning, and investigate the main cause of the fall of the elderly. 2. According to the software and hardware technology used to carry out the program design of the elderly anti-fall warning, an adjustable system is produced. 3. Design a series of experiments to test and correct the anti-drop system, use pre-recorded videos for the system to read, simulate the situation of real-time detection, and record and analyze the experimental data, and adjust and compare according to the data. 4. Based on the results of previous experiments, conduct a multi-regional integrated detection experiment to simulate the resilience of the system under the common bedroom decoration conditions in general houses.