Aerial Image Target Detection Algorithm Based on Improved CenterNet

Fig. 6. Aerial image target detection effect display. (a) Structured information dropout; (b) false and missed detection targets construct new samples; (c) night scene; (d) strong light scene; (e) (f) gathering area

Download full size

Table 1. Comparison of detection effects of different backbone networks
View table
Table 1. Comparison of detection effects of different backbone networks
Backbone mAP /% AP⁵⁰ /% AP⁷⁵ /% FPS /(frame·s^-1)
ResNet-50 20.05 42.05 19.75 65
DLA-34 22.50 45.88 20.50 55
ResDcn-18 14.01 36.00 15.25 131
Ours 25.22 45.62 23.36 45

Table 2. Comparison of effectiveness after successively introducing the designed structure
View table
Table 2. Comparison of effectiveness after successively introducing the designed structure
Backbone mAP /% AP⁵⁰/% AP⁷⁵ /%
ResNet-50 20.05 42.05 19.75
Ours+ (D-ASPP) 22.57 43.63 20.97
Ours++ (CBA-connection) 24.41 45.71 22.13
Ours+++ (Threshold) 25.22 45.62 23.36

Table 3. Evaluation of the recognition accuracies of different target types in the NWPU VHR-10 data set
View table
Table 3. Evaluation of the recognition accuracies of different target types in the NWPU VHR-10 data set
Target type AP /% Target type AP /%
Airplane 29.77 Basketball court 17.69
Ship 26.94 Ground track field 30.64
Storage tank 32.32 Harbor 29.31
Baseball diamond 31.29 Bridge 17.16
Tennis court 16.18 Vehicle 20.90

Table 4. Comparison of the detection effects of different detection algorithms on the NWPU VHR-10 data set

View table

Table 4. Comparison of the detection effects of different detection algorithms on the NWPU VHR-10 data set

Method	Backbone	mAP /%	AP⁵⁰ /%	AP⁷⁵ /%	FPS /(frame·s^-1)
Faster R-CNN	ResNet-101	27.23	48.57	24.50	9
YOLOv3	DarkNet-53	22.49	45.22	19.50	41
RetinaNet	ResNet-50	16.88	30.26	16.05	19
CornerNet	Hourglass-104	19.14	39.90	17.79	22
Ours	ResNet⁺	25.22	45.62	23.36	45

Tools

Get Citation

Copy Citation Text

Yanlei Xu, Jiran Liang, Guojun Dong, Zhuang Chen. Aerial Image Target Detection Algorithm Based on Improved CenterNet[J]. Laser & Optoelectronics Progress, 2021, 58(20): 2010013

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Image Processing

Received: Dec. 2, 2020

Accepted: Jan. 6, 2021

Published Online: Oct. 13, 2021

The Author Email: Jiran Liang (liang_jiran@tju.edu.cn)

DOI:10.3788/LOP202158.2010013

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Comparison of detection effects of different backbone networks

Table 1. Comparison of detection effects of different backbone networks

Table 2. Comparison of effectiveness after successively introducing the designed structure

Table 2. Comparison of effectiveness after successively introducing the designed structure

Table 3. Evaluation of the recognition accuracies of different target types in the NWPU VHR-10 data set

Table 3. Evaluation of the recognition accuracies of different target types in the NWPU VHR-10 data set

Table 4. Comparison of the detection effects of different detection algorithms on the NWPU VHR-10 data set

Table 4. Comparison of the detection effects of different detection algorithms on the NWPU VHR-10 data set