Small Scale Invade-Target Recognition and Location Based on Improved Faster RCNN

In order to realize the recognition and location of dynamic small-scale intrusion targets with the video monitoring system in unattended substations, a fast neural network identification method based on improved Faster RCNN is proposed. In this method, the strong semantic features of the target samp...

Full description

Saved in:
Bibliographic Details
Main Authors: Jingyi MA, Haoyang CUI, Mingda ZHANG, Yihui SUN, Yongpeng XU
Format: Article
Language:zho
Published: State Grid Energy Research Institute 2021-03-01
Series:Zhongguo dianli
Subjects:
Online Access:https://www.electricpower.com.cn/CN/10.11930/j.issn.1004-9649.202006190
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850053427798736896
author Jingyi MA
Haoyang CUI
Mingda ZHANG
Yihui SUN
Yongpeng XU
author_facet Jingyi MA
Haoyang CUI
Mingda ZHANG
Yihui SUN
Yongpeng XU
author_sort Jingyi MA
collection DOAJ
description In order to realize the recognition and location of dynamic small-scale intrusion targets with the video monitoring system in unattended substations, a fast neural network identification method based on improved Faster RCNN is proposed. In this method, the strong semantic features of the target samples are calculated by constructing the deep convolution network, and the location information is fused using the densely connected transmission channels, so as to obtain the basic backbone network suitable for small target detection. Then, the candidate region of the target is generated with the region proposal network, and the coordinates of the location frame are calculated using the bilinear interpolation method to achieve the accurate positioning at the pixel level. The model is trained based on the actual image sample set, and the improved Faster RCNN detection model is obtained. The experimental results show that the improved method can maintain high accuracy and timeliness in detection of small-scale foreign objects, and has a certain value for engineering application.
format Article
id doaj-art-c0e797939c0344149a2c1e3676661eef
institution DOAJ
issn 1004-9649
language zho
publishDate 2021-03-01
publisher State Grid Energy Research Institute
record_format Article
series Zhongguo dianli
spelling doaj-art-c0e797939c0344149a2c1e3676661eef2025-08-20T02:52:31ZzhoState Grid Energy Research InstituteZhongguo dianli1004-96492021-03-01543384410.11930/j.issn.1004-9649.202006190zgdl-54-02-majingyiSmall Scale Invade-Target Recognition and Location Based on Improved Faster RCNNJingyi MA0Haoyang CUI1Mingda ZHANG2Yihui SUN3Yongpeng XU4Department of Electronic and Information Engineering, Shanghai University of Electric Power, Shanghai 200090, ChinaDepartment of Electronic and Information Engineering, Shanghai University of Electric Power, Shanghai 200090, ChinaFenghua Power Supply Company, State Grid Zhejiang Electric Power Company, Ningbo 315500, ChinaFenghua Power Supply Company, State Grid Zhejiang Electric Power Company, Ningbo 315500, ChinaDepartment of Electrical Engineering, Shanghai Jiao Tong University, Shanghai 200240, ChinaIn order to realize the recognition and location of dynamic small-scale intrusion targets with the video monitoring system in unattended substations, a fast neural network identification method based on improved Faster RCNN is proposed. In this method, the strong semantic features of the target samples are calculated by constructing the deep convolution network, and the location information is fused using the densely connected transmission channels, so as to obtain the basic backbone network suitable for small target detection. Then, the candidate region of the target is generated with the region proposal network, and the coordinates of the location frame are calculated using the bilinear interpolation method to achieve the accurate positioning at the pixel level. The model is trained based on the actual image sample set, and the improved Faster RCNN detection model is obtained. The experimental results show that the improved method can maintain high accuracy and timeliness in detection of small-scale foreign objects, and has a certain value for engineering application.https://www.electricpower.com.cn/CN/10.11930/j.issn.1004-9649.202006190small-scale object detectiondeep learningconvolution neural networkfaster rcnnbilinear interpolation
spellingShingle Jingyi MA
Haoyang CUI
Mingda ZHANG
Yihui SUN
Yongpeng XU
Small Scale Invade-Target Recognition and Location Based on Improved Faster RCNN
Zhongguo dianli
small-scale object detection
deep learning
convolution neural network
faster rcnn
bilinear interpolation
title Small Scale Invade-Target Recognition and Location Based on Improved Faster RCNN
title_full Small Scale Invade-Target Recognition and Location Based on Improved Faster RCNN
title_fullStr Small Scale Invade-Target Recognition and Location Based on Improved Faster RCNN
title_full_unstemmed Small Scale Invade-Target Recognition and Location Based on Improved Faster RCNN
title_short Small Scale Invade-Target Recognition and Location Based on Improved Faster RCNN
title_sort small scale invade target recognition and location based on improved faster rcnn
topic small-scale object detection
deep learning
convolution neural network
faster rcnn
bilinear interpolation
url https://www.electricpower.com.cn/CN/10.11930/j.issn.1004-9649.202006190
work_keys_str_mv AT jingyima smallscaleinvadetargetrecognitionandlocationbasedonimprovedfasterrcnn
AT haoyangcui smallscaleinvadetargetrecognitionandlocationbasedonimprovedfasterrcnn
AT mingdazhang smallscaleinvadetargetrecognitionandlocationbasedonimprovedfasterrcnn
AT yihuisun smallscaleinvadetargetrecognitionandlocationbasedonimprovedfasterrcnn
AT yongpengxu smallscaleinvadetargetrecognitionandlocationbasedonimprovedfasterrcnn