Numeric Displays Detection System based on LabVIEW

The Solution:

We divided the project into three tasks:

First, processed the raw image into a uniform mode to reduce the influence of the illumination and LED indicators on the instrument. We utilized a series of techniques like brightness extraction and gamma collection to improve the contrast of the image (Fig1).

Second, located the position of the numeric displays. We calculated the energy center of the image, which implied the center of the displays after the first step, and detected four edges around center with particle analysis and linear regression. Based on the regression result, we got the position and angle of the displays. Next, we cut irrelevant parts and rotated the displays until it aligned with the horizon.

Finally, detected the numbers on the display with OCR method.

My Contribution:

1. I detected the numbers in a static image successfully by training a OCR detection.

2. In experiments, I noticed that the OCR failed when the numbers weren't aligned in a horizontal direction and some LEDs around the displays influenced the detection accuracy, so I designed an algorithm to detect the angle of the displays and cut the redundant part of image . However this method was not very robust, which was improved by Xiao later.

3. I found that the detection systems was influenced by the illumination conditions seriously. So I developed a algorithm to enhance the contrast of the image, shown in Fig 2.