Classification and recognition of emergency vehicles in traffic surveillance cameras offer an early warning to ensure the prompt reaction to the traffic about emergency stating. Much existing traffic surveillance camera detection monitoring based on computer vision technology has achieved high detection rates. Some of them are very innovative and have a higher novelty index. One of the innovative ideas is detecting an emergency vehicle in the moving traffic routes from the input of a video surveillance camera. Experiments conducted on various benchmarking databases show that the proposed algorithm successfully distinguishes the emergency vehicles from the traffic in a surveillance camera. The results show that the proposed algorithm has an innovative classification method, a higher accuracy rate, as well as comparable recall and specificity, compared with other methods. This paper proposes an algorithm that addresses a novel visual analysis technique for the detection of a moving emergency vehicle in traffic surveillance camera using the fundamentals of convolution neural network. Experiments conducted on several benchmark databases show that the proposed algorithm successfully distinguishes emergency vehicles from a test image. The results show that the proposed algorithm has an innovative classification method, higher accuracy rate, as well as comparable recall and specificity, compared to other methods. We created a very simple artificial neural network in Python with Keras and Tensorflow as backend, we learned to use a Python development environment with Anaconda and Jupyter Notebook, we also used numpy to handle arrays. From Keras we imported the Sequential model type and the "normal" Dense layer type. Finally we evaluated and predicted obtaining satisfactory results.