| Popis: |
In this paper, we present an efficient solution for automaticdetection and reading of dangerous goods plates ontrucks and trains. According to the ADR agreement dangerousgoods transports are marked with an orange platecovering the hazard class and the identification number forthe hazardous substances. Since under real-world conditionshigh resolution images (often at low quality) have tobe processed an efficient and robust system is required. Inparticular, we propose a multi-stage system consisting ofan acquisition step, a saliency region detector (to reducethe run-time), a plate detector, and a robust recognitionstep based on an Optical Character Recognition (OCR). Todemonstrate the system, we show qualitative and quantitativelocalization/recognition results on two challenging datasets. In fact, building on proven robust and efficient methods,we show excellent detection and classification resultsunder hard environmental conditions at low run-time.detection and reading of dangerous goods plates ontrucks and trains. According to the ADR agreement dangerousgoods transports are marked with an orange platecovering the hazard class and the identification number forthe hazardous substances. Since under real-world conditionshigh resolution images (often at low quality) have tobe processed an efficient and robust system is required. Inparticular, we propose a multi-stage system consisting ofan acquisition step, a saliency region detector (to reducethe run-time), a plate detector, and a robust recognitionstep based on an Optical Character Recognition (OCR). Todemonstrate the system, we show qualitative and quantitativelocalization/recognition results on two challenging datasets. In fact, building on proven robust and efficient methods,we show excellent detection and classification resultsunder hard environmental conditions at low run-time. |
