Detection And Recognition Of Traffic Signal Lamp Based On Dynamic Scene

This paper was based on automatic vehicle(auxiliary driving system) as the research object, detection and recognition of the traffic lights by using on-board visual was reseached, This paper includes the following aspects: 1. Traffic lights detection algorithm under dynamic scenariosThis paper was based on visual selective attention mechanism, and the human visual system was simulated to carry on selective attention to the focus information. That is, candidate target traffic lights was oriented with combination of the contrast, color, brightness and saturation, etc., and with the fusion of the local depth information, and then the candidate target is rapidly labeled by using run-length labeling algorithm, lastly, the interested targets of the candidate region was cut out. 2. the feature extractionTake shape features of cut arrow traffic lights and wavelet decomposition dimension reduction features of cut digital traffic lights for information extraction respectively. 3. The recognition of traffic lightsAfter the cut image feature extraction,the traffic lights need to identify, so we used binary forks tree method to classify the different traffic lights and then template matching algorithm was performed for the recognition of the traffic lights.Experimental results show that the method was improved of the existing traffic signal recognition method on the speed and robustness.