Offset calculation method for high and steep bedding slope visual displacement monitoring

Abstract

The invention relates to the technical field of visual monitoring, and discloses a high and steep bedding slope visual displacement monitoring offset calculation method, which comprises the following steps: S1, a camera arrangement and calibration step; the method comprises the following steps: S1, carrying out synchronous calibration and parameter matching on a main camera and an auxiliary camera, S2, carrying out initial image acquisition and processing, S3, carrying out real-time image acquisition and analysis, S4, carrying out initial offset calculation, and S5, carrying out offset calibration. According to the method, the built convolutional neural network is subjected to convolution, activation and pooling operation and is matched with stochastic gradient descent training, so that the feature points of the solar panel and the slope can be accurately identified, and the method is suitable for various environments; an interference area is efficiently detected through mahalanobis distance based on a CNN anomaly detection algorithm, and the risk is early warned in advance; and the generative adversarial network uses a generator to confront with a discriminator, so that high-quality image restoration is realized, the feature point matching and offset calculation precision is improved, and the monitoring accuracy and reliability are comprehensively guaranteed.

Images