| 摘要: | 眾所皆知，在羽毛球訓練或比賽後，場地上常會遺留大量的羽毛球，撿拾羽毛球並且排除嚴重損壞的羽球是非常耗費時間與人力的，在現在高度自動化的時代，應該有一套系統來協助運動員完成此任務；因此，本研究旨在設計一套系統，包含一款羽球撿拾機器人與羽球羽毛好壞檢測系統，其中羽球撿拾機器人可以自主辨識羽球場上散落的羽球，自動前去撿拾並收集；羽球羽毛好壞檢測系統則透過機械手臂，撿拾羽球至檢測區，經攝影機拍攝羽球，由我們發展的一套演算法檢查羽毛球的羽毛受損程度，判斷羽球優劣，並由運輸帶分類過濾掉過度受損的羽球。
整套系統分別由一台搭載YOLOv8影像實時辨識的撿球機器人與一個機械手臂以及分類運輸帶組成。整體運作流程如下，使用者可以透過手機啟動撿球機器人的自主撿球模式，機器人在球場自主偵測羽球位置並前往撿拾羽球，撿球任務完成後，機器人發送訊號至手機通知使用者，使用者可以透過機器人的遙控模式，控制機器人回到使用者身旁。在實際場地的實驗中，在半場羽毛球球場中，均勻灑落20顆羽球在場地中，分散面積約分布於長6.7公尺，寬6.1公尺的範圍內，機器人平均可將場地上92.79%的羽球撿拾，約需 134秒即可完成撿拾任務。在運行 YOLO 模型的情況下，機器人仍能達成即時效率。
將收集完的羽球放置在羽球檢測區後，機械手臂會將羽球夾取至攝影機前，攝影機拍攝的畫面回傳輸至電腦進行分析羽球的完整度，並將結果傳輸給Arduino控制分類運輸帶進而完成羽球品質好壞分類的任務。羽球好壞辨識的演算法中，使用影像處理，將羽球依照羽毛球羽毛數量，從羽球中心切分為16個區域，藉此來檢測羽毛球的羽毛是否為碎毛或羽毛缺失，而在此階段羽毛缺失較少的羽球，再藉由對羽毛邊緣形狀快速傅立葉轉換去分析羽毛邊緣的平整度，進而實現羽球好壞的判斷。
本系統提供使用者簡易的手機程式操作介面來控制羽球撿球機器人，並且將羽球好壞檢測的任務自動化，使用者僅需將羽球放置在檢測區，待分類任務完成後，即可獲得可以繼續使用的羽球。此系統若應用到實際運動場，將能有效的幫助運動員訓練時清理場地與分類羽球，簡化繁瑣耗時的工作。
;It is well known that after badminton training sessions or matches, a large number of shuttlecocks are often left scattered across the court. Collecting these shuttlecocks and discarding those that are severely damaged is a time-consuming and labor-intensive task. In today’s era of high automation, there should be a system to assist athletes in performing this task. Therefore, the aim of this research is to design an integrated system consisting of a shuttlecock collection robot and a shuttlecock quality inspection system.
The shuttlecock picking robot is capable of automated detecting and identifying scattered shuttlecocks on the court and then collecting them automatically. The shuttlecocks quality classification system utilizes a robotic arm to pick up the shuttlecocks and place them into a designated inspection area. An inspection camera captures images of each shuttlecock, and an algorithm developed in this research analyzes the degree of feather damage to determine its condition. The classification result is then used to sort out overly damaged shuttlecocks via a conveyor system.
The complete system consists of a YOLOv8-based real-time image recognition shuttlecock picking robot, a robotic arm, and a sorting conveyor belt. The operational workflow is as follows: the user can activate the mobile robot through a mobile application. The mobile robot automated searches the court, detects the location of shuttlecocks, and picks them up. Upon completing the collection task, it sends a signal to the user’s phone. The user can then remotely control the mobile robot to return using the mobile application. In field experiments, 20 shuttlecocks were randomly scattered over half of a standard badminton court, within an area of approximately 6.7 meters in length and 6.1 meters in width. On average, the mobile robot was able to collect 92.79% of the shuttlecocks in about 134 seconds, demonstrating real-time efficiency even while running the YOLO model.
Once the shuttlecocks are placed in the inspection area, the robotic arm transfers each shuttlecock to the front of the camera. The captured images are sent to a computer, which analyzes the shuttlecock’s condition. The results are then sent to an Arduino microcontroller that controls the conveyor system to complete the classification task. The shuttlecock inspection algorithm employs image processing techniques. Each shuttlecock is divided into 16 radial sectors from its center, based on the number of feathers. This helps identify whether the feathers are broken or missing. Shuttlecocks with minimal damage are further analyzed using Fast Fourier Transform (FFT) on the feather edge contour to evaluate the smoothness of the edge, ultimately determining the shuttlecock’s quality.
This system offers users a simple mobile interface to control the shuttlecock collection robot and automates the quality inspection process. Users only need to place shuttlecocks in the inspection area, and once the classification task is completed, the reusable shuttlecocks will be available.
If deployed in real-world sports environments, this system can effectively assist athletes in clearing the court and classifying shuttlecocks during training, significantly reducing repetitive and time-consuming tasks. |
