复杂环境下基于决策树的数码管识别算法研究

A DIGITAL TUBE RECOGNITION ALGORITHM BASED ON DECISION TREE IN COMPLEX ENVIRONMENT

目前基于图像的数码管自动识别方法在复杂的工业环境中存在模型失配、适应性不强、准确度较低的问题。为打通从理论研究到实际应用的路径,提升复杂环境下数码管识别的效率和性能,提出一种能适应复杂环境的、基于决策树的高效数码管识别算法。通过计算灰度图的Tsallis熵和均值,实现不同工作场景的适配,并在此基础上选择预设的算法完成灰度图二值化,利用数字定位算法完成数字区域的分割并进行识别。将该算法应用于实际环境中采集的大量图片,结果证明,该算法较好地适应了复杂工业环境,提供了更精准的识别结果。

At present,the image-based digital tube automatic recognition methods have the problems of model mismatch,poor adaptability and low accuracy in the complex industrial environment.In order to get through the path from theoretical research to practical application,and improve the efficiency and performance of digital tube recognition in complex environments,an efficient digital tube recognition algorithm based on decision tree that can adapt to complex environments is proposed.By calculating the Tsallis entropy and mean value of the grayscale image,the adaptation of different work scenes was realized,and on this basis,the preset algorithm was selected to complete the binarization of the grayscale image.The digital location algorithm was used to segment and recognize the digital region.The proposed algorithm was applied to a large number of pictures collected in the actual environment.The results show that th algorithm is well adapted to the complex industrial environment and provides more accurate recognition results.