台灣地區在各道路主管機關之調查作業上仍採用過去傳統的調查方 式，於車輛上採目視方式尋找破壞再進行人工量測，並在19 種常見的道 路破壞因子當中常常有現場工程師誤判的成因生成，這種方式不僅耗時耗 力，且於整體鋪面維護作業上耗費太多作業時間，本研究收集省道、快速 道路、市區道路與縣鄉道、國道高速公路等破壞態樣進行深度學習影像之 辨識。在硬體的選配與軟體的設計中，將設備安裝於道路巡查之車輛上， 以相機鏡頭擷取道路影像，透過網路將照片上傳至雲端，並再雲端自動進 行影像分析，辨識道路破壞並判讀進行道路分析及應用，以利收集大量鋪 面數據，並降低廠商之花費及人力資源之消耗。最後透過巡查機制來找出 破壞需養護之問題加以探討，以及作為即時性派工之應用。;In Taiwan, the traditional survey methods are still used in the survey operations of various road authorities. The vehicles are visually searched for damage and then manually measured. Of the 19 common road damage factors, there are often causes for misdiagnosis by field engineers. Generation, this method is not only time-consuming and labor-intensive, but also consumes too much operation time on the overall paving maintenance operation. This study collects damage patterns such as provincial roads, expressways, urban roads, county and rural roads, and national highways to carry out depth Learn image recognition. In the hardware selection and software design, the device is installed on the road inspection vehicle, the road image is captured by the camera lens, the photo is uploaded to the cloud through the network, and the cloud automatically performs image analysis to identify road damage It also interprets and analyzes roads and applications to facilitate the collection of a large number of paving data and reduce the cost of manufacturers and the consumption of human resources. Finally, through the inspection mechanism to find out the problems that need to be maintained and explore, and its application as an immediate dispatch.