基于深度学习改进算法的数字化变电站计量图像识别方法

Digital Substation Measurement Image Recognition Method Based on Improved Deep Learning Algorithm

摘要: 针对当前数字化变电站计量领域图像识别准确度有待提高的问题，本文提出一种基于深度学习改进算法的图像识别方法。该方法对传统的图像识别算法进行优化，使得合成的形状更接近目标对象，同时改进了卷积层，通过采用可变形卷积，加强对计量设备图像不同物体形变的建模能力。论文详细介绍了改进方法的原理、实现算法和实现流程，并开发了对应的计量图像识别软件，验证结果表明该算法具有良好适应性和高识别可靠度等优点。

Abstract: Aiming at the problem that the accuracy of image recognition in the field of digital substation measurement needs to be improved, this paper proposes an image recognition method based on improved deep learning algorithm. This method optimizes the traditional image recognition algorithm to make the synthesized shape closer to the target object, and improves the convolution layer. By using deformable convolution, the ability of modeling the deformation of different objects in the measurement equipment image is enhanced. This paper introduces the principle, algorithm and process of the improved method in detail, and develops the corresponding measurement image recognition software. The verification results show that the algorithm has the advantages of good adaptability and high recognition reliability.