# 2022 IOT # 路邊停車格即時資訊 ##### Real-time On-street Parking System for Scooters and Motorcycle 組員： 施育衡，周宗翰，劉柏顯，陳捷文，陳威宇 --- ## Motivation ---- <!-- 台灣的交通主要由機車所構成，並且多數機車族最大的煩惱莫過於尋找停車位，路邊的機車位成了他們的首選。 --> * Traffic mainly consisted of scooters and motorcycles * Problem of finding parking space * Resolve to on-street parking ---- <!-- 但是路邊的停車格與有管理的停車場不同，沒辦法很好的追蹤停車數量。 --> * Difference between on-street parking and parking lot * Difficult to track parked vehicle ---- <!-- 因此使用者無法預測此路段何時會有路邊車位，或此路段的停車壅擠程度。 --> <!-- 所以我們擬研發一套即時機車路邊停車系統， --> * No well-designed system for tracking on-street empty parking space * Developing a real-time on-street parking system ---- <!-- 方便使用者追蹤目的地附近車位，減少等待時間，進一步改善交通問題。 --> * Track on-street parking space * Minimize their waiting time * Improve overall traffic congestion. --- ## Feature ---- <!-- 紀錄、分析台灣路邊機車停車格的車位資訊。 --> 1. Record and analyze parking space information <!-- 提供使用者即時的車位資訊，並透過使用者的位置提供車位推薦。 --> 2. Provide real-time information and recommandation --- <!-- ## 商業價值 --> ## Commercial Value ---- <!-- 分析使用者喜好，並結合附近展覽，活動，或美食店家等, 推薦哪一些停車空位附近是否有使用者感興趣的事物。 --> <!-- 另外，分析目標停車場附近的交通概況，加速人流流動，增加曝光率與遊客人數。 --> 1. Analyzing user preferences 2. Combine with nearby exhibitions, events, and restaurants 3. Recommend parking spaces --- <!-- # 參考文獻與討論 --> Related Work === # and Discussion ---- ## Detection ---- <!-- Yolo是一個物體辨識的深度學習模型，給予一張影像，模型會辨識物體的類別與 bounding box 的位置。 --> ### YOLOv7 * Deep learning model for object detection * Categorize objects * Frame out bounding box <!-- 我們可以使用detection的方法經過辨識後，再對偵測後的label進行處理 --> ---- ### Usage * recognize vehicles * Process label. <!-- 優點: 可以同時汽車與摩托車的停車狀態 --> ---- ### Benefits * Accurately detect vehicle * Classify vehicle types <!-- 缺點: 若是影像過大或物體過於密集，此模型處理速度慢且性能不佳 --> ---- ### Drawbacks * Processsing time efficiency --- ## Counting ---- ### Car Counting with CNN <!-- 透過輸入影像，輸出影像的density map，再透過積分的方式數出圖中關注的物體數量 --> * Predict density map * Integral methods to count relevent objects. ---- ### Benefits <!-- 優點: 可以在物體密集及嚴重遮擋的情況下，很好地預測圖中的物體數量。 --> * detect objects from severly blocked and congested environment. ---- ### Drawback <!-- 缺點: 無法精準定位與得知物體大小 --> * cannot precisely locate objects position * cannot measure object size. --- ## Data Analysis ---- ### IoT Smart Parking System with LSTM * Time wasted for finding parking space * Predict empty parking space --- <!-- # 問題與挑戰 --> # Problem and Challenges ---- <div style="display: flex; flex-direction: column; justify-content: flex-start; text-align: left; font-size: 33px;"> <span>1. The need for short delay real-time system</span> <span>2. Rapid data update due to mass flow of traffic </span> <span>3. Detect vehicle and user intent.</span> <span>4. Trade off maximize area coverage and minimize cost.</span> <span>5. Interference to sensors by obstacles and extreme condition</span> <span>6. Wrong vehicle on parking slots </span> <span>7. Congested parking space analysis caused by illegal parking and insufficient space.</span> <span>8. Project cost should be at a fair price</span> <span>9. Commercial advertisement to user combined with parking.</span> </div> --- # Sensors ---- * Camera (sigular or multiple): Webcam or CCTV camera (differ from video quality and price) * Ultrasonic sensors * Radar sensors --- # Reference - [YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors](https://arxiv.org/abs/2207.02696) - [An accurate car counting in aerial images based on convolutional neural networks](https://link.springer.com/content/pdf/10.1007/s12652-021-03377-5.pdf) - [IoT Based Smart Parking System Using Deep Long Short Memory Network](https://www.mdpi.com/2079-9292/9/10/1696/htm) --- Thanks for Listening