Aiming at special environmental requirements of traffic signal detection task

such as car tail light

background variability

lighting problem

etc

combined with conventional neural network（ CNN）

put forward random point as the center

and gradually spread

while according to features of context of different sizes detect context feature detection algorithm of each step of the results of the spread. Each step of the spread will cover the context information around the target

with the traffic lights has context information will be easier to distinguish with the car taillights. According to the experimental analysis

compared with other algorithms

it improves recall rate

at the same time

can effectively avoid false inspection of car taillight

and solve the problem of low recall rate and high false alarm rate in traffic signal detection.