| 摘要: | 「人流」在各領域皆具有重要性與應用價值。在生產管理方面，透過人流可以評估生產線效率、發現瓶頸、優化布局、提高生產效率和品質。在行銷管理中，人流可用於評估顧客流量、消費者行為，指導陳列和促銷策略，提高銷售額和客戶滿意度。在人力資源管理中，它有助於預測人力需求，調整員工配置，改善工作環境，提高員工滿意度和工作效率。在財務管理方面，人流用於評估營業時間的利用率，控制成本，提高經營效率，增加收入。總之，人流成為企業提升生產效率、行銷策略、人力資源配置和財務管理的關鍵工具，促進企業永續發展。
員工餐廳在食材、成本和供應商管理方面遇到困難。難以準確預測麵食類和飯菜類的需求量，可能導致食材浪費或供應不足，增加成本。同時，即時領餐人流可以提供當前的備菜量是否足夠，是否需要增加解決可能導致庫存積壓和供餐品質問題，影響供應商關係。缺乏即時人流計數也將影響與供應商的合作，難以精確反映需求，可能導致訂單錯誤和庫存管理問題。
這次研究利用CSI Camera捕捉影像，傳送至Jetson Orin Nano進行即時影像處理，搭配YOLOv8進行人物檢測及追蹤，再透過RESTful API將結果以web API方式提供其他系統取得人數數據，實現高效即時人流計數。
人流計數系統可以滿足營養科在領餐人流計數，不過在這次的研究中發現需要有環境控制上的配合才可以逹到更精準的人流計數，在研究過程中因YOLOv8的人物偵測及追蹤無法滿足計數人數的需求，所以我加上了使用基準線的方式來計數人流，在實際上線後發現需要有環境上的控制才可以逹到精準的人數計數功能，在人流計數系統上的數字也可以讓醫院的儀表板取得當下的領餐人數，營養科同仁可以依據歷史數據評估當前用餐人數是否合理，進而評估目前的備菜量是否足夠、是否需做調整，在用餐成本支出部分，因為有了麵食類及飯菜類的領餐人數，所以當成本支出異常時可以協助快速分辨是麵食類或飯菜類領用異常，不在像以往只有指紋機的總額數據。
;"Foot traffic" holds significant importance and application value across various fields. In production management, analyzing foot traffic helps assess production line efficiency, identify bottlenecks, optimize layouts, and enhance both productivity and quality. In marketing management, it is used to evaluate customer flow and consumer behavior, guide merchandise display and promotional strategies, thereby increasing sales and customer satisfaction. In human resources management, it aids in forecasting manpower needs, adjusting staff allocation, improving work environments, and boosting employee satisfaction and work efficiency. In financial management, foot traffic is used to evaluate the utilization of business hours, control costs, improve operational efficiency, and increase revenue. Overall, foot traffic has become a key tool for businesses to enhance production efficiency, marketing strategies, human resource allocation, and financial management, promoting sustainable development.
The staff cafeteria faces difficulties in managing ingredients, costs, and suppliers. The challenge of accurately predicting the demand for noodle and rice dishes can lead to either wastage of ingredients or insufficient supply, increasing costs. Real-time foot traffic data during meal pickup can indicate whether the current preparation is sufficient or if adjustments are needed to address potential stockpile and meal quality issues, impacting supplier relationships. A lack of real-time foot traffic counting also affects cooperation with suppliers, as it is difficult to accurately reflect demand, potentially leading to order errors and inventory management issues.
This study employs a CSI Camera to capture images, which are then processed in real time by a Jetson Orin Nano, combined with YOLOv8 for person detection and tracking. The results are then made available through a RESTful API via a web API method for other systems to access the foot traffic data, achieving efficient and real-time foot traffic counting.
The foot traffic counting system can satisfy the needs of the nutrition department in counting meal pickups. However, this study found that environmental control is necessary to achieve more accurate foot traffic counts. Due to the inadequacy of YOLOv8′s person detection and tracking for counting, I incorporated a baseline method to count foot traffic. Upon deployment, it was found that environmental control is necessary to achieve precise foot traffic counting. The numbers from the foot traffic counting system can also provide the hospital dashboard with current meal pickup counts, allowing nutrition department colleagues to use historical data to evaluate whether current dining numbers are reasonable, assess if the current food preparation is sufficient, and make necessary adjustments. In terms of dining cost expenditure, having the counts for noodle and rice dish pickups helps quickly identify if there are any anomalies in the costs of these specific categories, unlike in the past when only total data from fingerprint machines was available. |
