Our team developed a deep learning algorithm for sequential signal recognition. The signal feature will be extracted by the convolutional neural network, and the time state will be inferred through the long short-term memory. Final inference on the state of the parking space.

為實現路邊停車管理服務，我們採用地磁感應器於空位狀態分析，並同時規劃以長距離低功耗物聯網傳輸技術(LoRa)當成物聯網通訊，但在實際場景中，地磁感測器易受外在環境變化之影響。例如環境磁場、外在量測訊號雜訊等等。所以我們團隊開發適合於時序性訊號樣式辨識之深度學習演算法。將藉由卷積神經網路(CNN)提取地磁訊號特徵，在通過長短期記憶模型(LSTM) 推論時間軸空位狀態，不斷觀測和記憶停車狀態，並對停車位狀態做出最終推斷。以提升辨識率與穩定度，達到實現路邊車格之停位管理。

We will promote the parking management system to the daily life circle, such as department stores. By connecting the system, vehicles and users through the Internet of Things, a more user-friendly parking management platform and payment mechanism can be provided.