Apparent characteristic monitoring data intelligent labeling method based on target detection

Abstract

The invention discloses an apparent characteristic monitoring data intelligent labeling method based on target detection. The apparent characteristic monitoring data intelligent labeling method comprises the following steps: acquiring and processing data; marking the acquired data; carrying out preliminary training on the model; detecting the new data; importing the existing annotation data and the model training result thereof into a data annotation system; finely adjusting a result, detecting the classified and trained data, and correcting an error result; and iterative optimization is performed on the model, the adjusted data are used for iterative optimization of the model, the learning rate eta and regularization parameter lambda hyper-parameter calculation in the neural network modelin the model is adjusted until the output error reaches the final requirement, and training is ended. According to the method, the time for apparent characteristic target detection and data annotation in various fields of slope and landslide disaster prevention and control and weir dam disaster treatment is saved, the efficiency and accuracy of data annotation are improved, and the method conforms to intelligent application of various technologies of big data storage and cloud calculation in the fields of engineering construction and disaster prevention and control.

Description

technical field [0001] The invention relates to frame rate extraction in image processing, image noise reduction, and a target detection and neural network model training method in the field of artificial intelligence, in particular to an intelligent labeling method for monitoring data of apparent features based on target detection. Background technique [0002] At present, all walks of life are developing towards artificial intelligence methods. Target detection and data labeling play a vital role in the development of artificial intelligence. In the field of concrete crack detection, a large number of target detection and data labeling are required. [0003] The task of target detection is to find all the required target objects in the image and determine their position and size, which is one of the core problems of machine vision. [0004] With the development of the trend of the times, the fields of major project construction and disaster prevention and control are acti...

AI Technical Summary

Problems solved by technology

However, at the current stage, there is a serious lack of labeled data in the process of intelligentization of major projects

Moreover, the existing data labeling systems usually require a lot of manpower and time investment, and when labeling data, a large number of professionals are also required. For example, the target detection in the medical industry must be carried out by doctors; in the field of civil engineering, landslides In the process of apparent displacement monitoring or underground cavern safety maintenance monitoring, the generation and development of some deformations and cracks need to be identified and marked by practitioners, which is inefficient and error-prone

Images