A method for detecting irregularities in electric power smart construction sites based on improved yolov4 algorithm

Abstract

The invention discloses a method for detecting irregularities in electric power smart construction sites based on the YOLOv4 improved algorithm. Collect the image of the electric power construction site used for training the model; perform image enhancement on the collected image of the electric power construction site; mark the target area obtained after image and data enhancement with a rectangular frame, and obtain the coordinates of the rectangular frame and the According to the image of the construction site collected and the image obtained after data enhancement, and the obtained coordinates of the rectangular frame and the types included, the improved model based on YOLOv4 is trained; the image of the electric power construction site is collected in real time, according to The trained model collects real-time images of power construction sites to be detected, and outputs images of violations. This method is applicable to the detection of personnel violations and construction vehicle violations on the construction site, and realizes the visible, precise and intelligent management of the electric power smart construction site, effectively improving the management level of the project site and reducing safety risks.

Description

technical field [0001] The invention is applicable to the field of electric power smart construction site monitoring, and relates to a method for detecting personnel and vehicles, in particular to a method for detecting irregularities in electric power smart construction sites based on the YOLOv4 improved algorithm. [0002] technical background [0003] In order to strengthen project site management and control, it is necessary to strengthen the monitoring and detection of personnel and vehicles, deepen video recognition and video analysis capabilities, and improve supervision efficiency. In recent years, with the rapid development of deep learning technology, the Object Detection algorithm has also shifted from the traditional algorithm based on manual features to the detection technology based on deep neural network. YOLO is an effective object detection model without region proposals, which can directly train the entire network end-to-end, and its main feature is the obvi...

AI Technical Summary

Problems solved by technology

[0004] At present, YOLO-based detection methods are all based on YOLOv3 to build a network model, such as a face detection method, device and equipment based on convolutional neural network (Chinese patent, application number CN202010058934.6, publication date: 2020-06-16) , a YOLOv3-based aerial drone target recognition and tracking method (Chinese patent, application number CN201911394465.9, publication date: 2020-06-05), a safety helmet wearing detection method and device based on deep learning (China Patent, application number CN201911349221.9, publication date: 2020-06-16), these detection models are all based on YOLOv3 to build a network model, the detection speed and accuracy will be inferior to YOLOv4

Images