療養膳食自動化品檢技術－－應用於醫院中央廚房

Automated Dietary Inspection Techniques for Hospital Central Kitchen Application

本研究建構了一套統應用於醫院中央廚房之自動化膳食檢測系統。為符合醫院之病患之個人化營養需求，如少油、低鹽等特殊營養攝取目的，過去皆是由醫院營養部之營養師逐一手動檢視，傷神費力。在本文所提出之系統中，藉由視覺處理之方法，自動化定位在輸送帶上製作之膳食餐盒；其次，將餐盒上所乘載之膳品透過色彩（polar histogram）及紋理（LBP-HF）之特徵來進行品樣識別；最後，針對各膳品之份量利用3D深度資訊進行膳品份量識別。實驗結果中，膳食餐盒之識別精度80%以上，其每個膳食餐盒之品樣及份量檢測速率可達1500毫秒以內及其膳品識別率在80%左右。預期本系統以自動化檢測方法來提升中央廚房之出餐量50%以上，並有效減低營養師進行人工全檢之品管壓力及作業時間。

This paper introduced an intelligent and automatic dietary inspection system which could be applied to the dietary preparation at the hospital central kitchen. The diet prescriptions specifically designed for patients with various disease diagnoses are required as low sodium intake or caution with some food. Hence, the proposed system could benefit the inspection process that has been performed manually by nutritionist. The meal box would be detected and located automatically with the vision-based method and then all the food ingredients could be identified by using color and the LBP-HF texture features. Furthermore, the quantity of each food ingredient would be estimated by using the image depth information. The result showed that the dietary inspection accuracy was around 80%, dietary inspection efficiency had reached 1500 ms, and the food quantity accuracy was about 75% for per meal box. The proposed system is expected to increase the capacity of meal supply over 50% by saving the time in the diet inspection process in the hospital.