Image data processing method and device based on multi-target detection

Abstract

The invention discloses an image data processing method and device based on multi-target detection. The method comprises the steps that an image is acquired, the image is preprocessed, image features are extracted from a network through basic convolution, and a feature map of the image is obtained; generating a large number of default suggestion boxes for the feature map through a pre-constructed regional suggestion network, and classifying the feature map and the suggestion boxes; based on an area-based non-maximum suppression algorithm, performing de-duplication on the classified suggestion box to finish processing, and improving the brightness and contrast of the image and the adaptability to a complex environment image through a self-adaptive preprocessing algorithm; the A-NMS algorithm based on the area is used for solving the problem of single-target multi-detection frames so as to improve the discrimination capability of the algorithm on tiny feature differences; and a segmentation detection method is adopted, so that the recognition capability of the algorithm on small targets is improved, the omission ratio of long-range targets is reduced, and more accurate fault detection on multiple targets in the aerial image under a complex background is obtained.

Description

technical field [0001] The invention relates to an image data processing method and device based on multi-target detection, belonging to the technical field of data processing. Background technique [0002] Power security is the guarantee for the smooth operation of the national economy. Since the transmission line has been exposed to the wild for a long time and the environment is harsh, regular inspections of the transmission line are required to find faulty or hidden parts and replace them in time. At present, the traditional manual line inspection method has been gradually replaced by drones and surveillance cameras. Electric workers do not need to visit the line in person, but only need to identify pictures in the monitoring room. In the face of a large number of returned images, purely manual identification will consume a lot of manpower and energy. Therefore, how to use computer technology to realize automatic identification of faults is a hot topic for current resea...

AI Technical Summary

Problems solved by technology

Wang Wanguo et al. used UAV inspection images to test the effects of various target detection algorithms on component positioning and classification with common electrical components as targets, and achieved a correct rate of 92.7%, but failed to achieve fault discrimination.

Images