Abstract:  In order to solve the problem that the spray image acquired through visual fuel injection image
acquisition system is low contrast and the boundary is not clear. A binarization algorithm of adaptive threshold
value based on equalization histogram is presented. According to the histogram properties of the aerosol image
enhanced by contrast enhancement, the algorithm can get the adaptive binarization threshold. According to the
binarization result, the image contour is extracted, the nozzle position, spray image outermost tangent, and spray
farthest point position can also be detected automatically while the two characteristic parameters of spray image
the spray penetration and spray angle can be calculated. 25 images collected in a whole process of fuel injection
are processed. The experimental results show that the method can extract the contour of the spray image more
clearly and accurately. The trend of change in the spray penetration and the spray angle is in accordance with the
theoretical value.

Key words: feature extraction, spray angle,  image processing, spray penetration