電腦視覺應用於車輛檢驗站中車輛尺度量測之研究

Abstract

隨著社會大眾對車輛安全的意識抬頭，各國持續改善道路及車輛設計，加強法規制定及執法效率。而本國車輛檢驗於車輛尺度量測工作中，仍依賴於人工方法，導致檢驗單位存在消耗人力、耗時、量測誤差和職業風險等負面影響。 本研究為優化檢驗工作之效率，以各種感測器之量測技術進行全面回顧及比較評估，選擇光學圖像感測器 (CMOS) 作為車輛尺度量測之感測器，裝設於本國車輛檢驗站之車道，並採用多目視覺方法，通過影像處理演算法設計，對車輛影像擷取之像素特徵進行析，量測車輛尺度。 通過實驗驗證評估，以電腦視覺能有效量測車輛尺度，結果顯示系統於速度、準確性、穩定性及減少量測誤差方面優於傳統人工量測方法，可作為車輛尺度量測之參考。

With the increasing public awareness of vehicle safety, the whole world Continuously improved road and vehicle design and enhanced the formulation and enforcement of regulation. However, vehicle dimensions measurement still relies on manual methods in Taiwan, leading to negative impacts such as manpower consumption, time consumption, measurement errors, and occupational risks. To optimize the efficiency of the inspection process, this research conducted a comprehensive review and comparative evaluation of various sensor measurement technologies. Optical image sensors (CMOS) were selected as the perceptual sensors for vehicle dimensions measurement. They were installed in the lanes of Taiwan’s vehicle inspection stations. Multicamera stereo vision methods were employed, and an image processing algorithm was designed to analyze the pixel features of captured vehicle images and measure vehicle dimensions. Through experimental verification and evaluation, it was confirmed that computer vision can effectively measure vehicle dimensions. The results showed that the system outperforms traditional manual measurement methods in terms of speed, accuracy, stability, and reduction of measurement errors. It can serve as a reference for vehicle scale measurement.