Keycap character detection model based on target classification
Tian‐Jian Xiong, Zhiguo Zhu, Tinghao Wang · Journal of Physics Conference Series · 2024
Abstract The production process of keycaps may result in surface character defects. The existing keycap character defect detection technology has low efficiency and accuracy, hindering the automation of keycap manufacturing. Addressing the challenge of extracting the entire keyboard from illuminated regions, this paper proposes a method using image feature points for template matching to extract the region of interest. Subsequently, leveraging prior knowledge, the paper addresses character segmentation and extraction on the keyboard. For character detection, the paper preprocesses images and classifies them by using an S-VGG network, yielding corresponding labels.