Image Recognition Algorithm of UAV Inspection under Machine Vision

UAVs have the advantages of efficient and automated inspection in life, and have important application value in industry, construction, energy and other fields. In this paper, an improved image recognition algorithm based on machine vision is proposed to solve the problems existing in UAV inspection. Through the use of computer vision technology, this paper analyzes and processes the images captured by unmanned aerial vehicles to complete the automatic recognition and defect detection of objects. A complete algorithm system composed of image preprocessing, feature extraction, object classification, and defect recognition is also proposed. Using large-scale drone patrol image data, the method is evaluated from three aspects: accuracy, F1 value and recognition speed. After testing, the prediction accuracy of this method can reach 89% to 96%. The algorithm has made significant improvements in target recognition and defect detection. The improvement in accuracy and F1 score indicates that the algorithm can identify the target.