III. MAIN ALGORITHM
A. Barcode location algorithm
Take the factors that high resolution, complex background
of electric meter image, this paper divide the process of
location into coarse positioning and fine positioning two parts.
Coarse positioning process the low resolution image via the
methods such as feature extraction, morphologic, connected
domain and so on. Fine positioning research the objects which
are the result of coarse positioning via the methods such as
edge extraction and gray projection. Experimental results show
that the proposed method have a good performance when the
captured image with complex background and non-uniform
lighting, at the same time, it can be running in embedded
system real-time.
B. Image binarization algorithm
A binarization method which global and local information
are used together is proposed in this paper. The global
threshold is got by the prior knowledge of barcode first, and
then the difference value of the mean of sub-regions and initial
image is calculated as the threshold range of sub-regions. At
last, the threshold is got using OTSU method in this range. The
experimental result shows that the performance of our
algorithm is well when the input image is non-uniformly.