Image　recognition　technology　based　on　convolutional　neural　network　(CNN)　has　been　widely　used　in　the　field　of　intelligent　transportation　in　recent　years.　Since　the　image　recognition　in　the　field　of　intelligent　transportation　needs　high　real-time　performance,　this　requires　improving　the　speed　of　CNN.　We　refer　to　Overfeat,　which　was　proposed　in　the　ImageNet　Large　Scale　Visual　Recognition　Challenge,　to　build　a　vehicle　and　pedestrian　recognition　model.　We　do　not　use　the　traditional　sliding　window　method.　Instead,　we　apply　each　convolution　over　the　extent　of　the　full　image,　eventually　producing　a　map　of　output　class　predictions.　This　method　ensures　the　accuracy　of　image　recognition,　while　enhancing　the　operational　efficiency　and　the　real-time　performance　of　CNN.　In　this　paper,　we　use　both　a　new　method　and　the　traditional　sliding　window　method　for　the　recognition　of　pedestrians　and　cars　on　the　road.　Then,　we　compare　the　advantages　and　disadvantages　of　the　two　methods　in　terms　of　their　recognition　effect　and　speed.　©　2015　ACM.