|
In traditional industries, there is also a need for intelligence to reduce or correct the more traditional methods of their factories, such as manual recognition, monitoring, and writing. This study focuses on generating an automatic recognition system for traditional industries to greatly improve the efficiency of personnel in monitoring and observing steel coil markings, thereby facilitating online production efficiency.
First, due to the unavailability of on-site information, this study uses the method of rephotographing to obtain images of the production line surveillance system. After dividing the obtained images into individual frames, the images are filtered and analyzed. Then, due to the different production times and minimal differences in image content, this study utilizes familiarity with on-site operation status and analyzes the internal composition of the images. It proposes the use of masks in fixed positions to make the image features more prominent without losing their status characteristics. This helps filter out lots of unnecessary images, such as moving steel coils, repeated steel coils, and unwritten identities. Finally, immediately transmit the filtering results to the Google Cloud Vision API for font recognition, and write the results into the image to display them to the user immediately. This allows the user to understand whether the font written by the current coil writing machine is correct and whether the font is normal.
This study provides a more convenient way for personnel to identify the identity of steel coils, with an average recognition rate of 91% during the day and 94.8% at night. However, due to adverse factors such as the on-site environment, the accuracy rate during the day and at night is only 11% and 55%.
|