One way to solve this problem is to select pixels in the road before the classification. I think this will help the algorithm to solucionate the problem of the misclassification.

Which algorithm are you using? maybe you can try one that are more robust too, like Random Forest!

Best regards

Roberto Filgueiras thnxx and is this available in ArcGis

This link shows how to perform random forest classification in ArcGis with Python. I did not use this methodology yet.

https://learn.arcgis.com/en/projects/predict-seagrass-habitats-with-machine-learning/lessons/perform-random-forest-classification.htm