PE-Net：一种优化剪枝的实时山体滑坡检测网络

doi:10.12305/j.issn.1001-506X.2025.08.06

Abstract

Abstract:

Real-time detection of slope landslides is crucial for reducing casualties and property damage. To address the issues of time lag and misjudgment in traditional object recognition methods for landslide monitoring, a multi-domain dataset is constructed to enhance the understanding of the visual features of slope landslides and sandstorms, and proposes a pruned slope and enhanced model named PE-Net for landslide automatic detection. This model is based on an improved VanillaNet network and incorporates a dynamic multi-head attention detection block, significantly enhancing the visual perception capability for landslide scenarios. Additionally, the model is compressed using the performance-aware approximation of global channel pruning （PAGCP） algorithm to facilitate embedded deployment. Experimental results demonstrate that the proposed model significantly improves the accuracy of slope landslide detection in real-time scenarios, providing valuable insights for natural disaster monitoring and warning of slope landslides.

Key words: landslide, object detection, VanillaNet, dynamic detection head, global channel pruning

CLC Number:

TN 911.73

Ying YU, Chunping WANG, Jinhui XU, Shuhang LYU, Qiang FU, Ming CHEN. PE-Net: a real-time landslide detection network with optimized pruning[J]. Systems Engineering and Electronics, 2025, 47(8): 2475-2485.

Figures/Tables 20

Fig.1

Fig.2

Fig.3

Fig.4

Fig.5

Fig.6

Table 1

Fig.7

Fig.8

Table 2

Fig.9

Fig.10

Fig.11

Table 3

Fig.12

Table 4

Table 5

Table 6

Table 7

Table 8

References 35

1	JIA H Y, WANG Y J, GE D Q, et al. Improved offset tracking for predisaster deformation monitoring of the 2018 Jinsha River landslide （Tibet, China）[J]. Remote Sensing of Environment, 2020, 247, 111899. doi: 10.1016/j.rse.2020.111899
2	ZHAO C Y, LIU X J, GAO Y, et al. Early identification of high-elevation geohazards in the lower Yarlung Zangbo River based on the SAR/InSAR technology[J]. Journal of Geomechanics, 2022, 26 (6): 981- 994.
3	ZHANG L F. Early identification of hidden dangers of loess landslide based on time series InSAR: a case study of southwest Bailuyuan[J]. Northwestern Geology, 2023, 56 (3): 250- 257.
4	许濒支, 朱庆, 李海峰, 等. 基于多尺度深度注意力模型的遥感影像滑坡提取方法[J]. 测绘地理信息, 2022, 47 (3): 108- 112.
	XU B Z, ZHU Q, LI H F, et al. A landslide extraction method of remote sensing image based on multi-scale depth attention model[J]. Journal of Geomatics, 2022, 47 (3): 108- 112.
5	YANG Y, YANG J T, XU C D, et al. Local-scale landslide susceptibility mapping using the B-GeoSVC model[J]. Landslides, 2019, 16 (7): 1301- 1312. doi: 10.1007/s10346-019-01174-y
6	NAPOLI M D, CAROTENUTO F, CEVASCO A, et al. Machine learning ensemble modelling as a tool to improve landslide susceptibility mapping reliability[J]. Landslides, 2020, 17 (8): 1897- 1914. doi: 10.1007/s10346-020-01392-9
7	陈雪梅, 刘志恒, 周绥平, 等. 基于HRNet的高分辨率遥感影像道路提取方法[J]. 系统工程与电子技术, 2024, 46 (4): 1167- 1173.
	CHEN X M, LIU Z H, ZHOU S P, et al. Road extraction from high-resolution remote sensing images based on HRNet[J]. Systems Engineering and Electronics, 2024, 46 (4): 1167- 1173.
8	CHENG G, GUO L, ZHAO T Y, et al. Automatic landslide detection from remote-sensing imagery using a scene classification method based on BoVW and pLSA[J]. International Journal of Remote Sensing, 2013, 34 (1): 45- 59. doi: 10.1080/01431161.2012.705443
9	AMATYA P, KIRSCHBAUM D, STANLEY T. Use of very high-resolution optical data for landslide mapping and susceptibility analysis along the Karnali Highway, Nepal[J]. Remote Sensing, 2019, 11 (19): 2284. doi: 10.3390/rs11192284
10	张居福. 自动化监测系统在滑坡变形监测中的应用研究[J]. 工程技术研究, 2023, 8 (15): 103- 105.
	ZHANG J F. Research on the application of automatic monitoring system in landslide deformation monitoring[J]. Engineering and Technological Research, 2023, 8 (15): 103- 105.
11	WANG Y T, GAO H, LIU S H, et al. Landslide detection based on deep learning and remote sensing imagery: a case study in Linzhi City[J]. Natural Hazards Research, 2025, 5 (1): 95- 108.
12	JI S P, YU D W, SHEN C Y, et al. Landslide detection from an open satellite imagery and digital elevation model dataset using attention boosted convolutional neural networks[J]. Landslides, 2020, 17 (6): 1337- 1352. doi: 10.1007/s10346-020-01353-2
13	CASAGLI N, INTRIERI E, TOFANI V, et al. Landslide detection, monitoring and prediction with remote-sensing techniques[J]. Nature Reviews Earth & Environment, 2023, 4, 51- 64.
14	胡邦国, 孙晓龙, 陈立, 等. 天地一体化地物变化监测与风险评估[J]. 中国矿业, 2020, 29 (S2): 135- 141.
	HU B G, SUN X L, CHEN L, et al. Automatic detection technology and system for ground feature changes and geological risk assessment along oil pipeline based on space-ground-joint sensing[J]. China Mining Magazine, 2020, 29 (S2): 135- 141.
15	LU J, GONG P H, YE J P, et al. A survey on machine learning from few samples[J]. Pattern Recognition, 2023, 139, 109480. doi: 10.1016/j.patcog.2023.109480
16	WANG Y Q, YAO Q M, KWOK J T,. Generalizing from a few examples: a survey on few-shot learning[J]. ACM Computing Surveys, 2020, 53 (3): 226931458.
17	ZHANG H Y, CISSE M, DAUPHIN Y N, et al. Mixup: beyond empirical risk minimization[C]// Proc. of the 6th International Conference on Learning Representations, 2018.
18	ZHONG Z, ZHENG L, KANG G L, et al. Random erasing data augmentation[EB/OL].https://arxiv.org/abs/1708.04896.
19	CHEN H T, WANG Y H, GUO J Y, et al. VanillaNet: the power of minimalism in deep learning[C]// Proc. of the 37th Conference on Neural Information Processing Systems, 2023.
20	DAI X Y, CHEN Y P, XIAO B, et al. Dynamic head: unifying object detection heads with attentions[C]// Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021: 7369−7378.
21	HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C]// Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2016: 770−778.
22	SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[C]// Proc. of the 3th International Conference on Learning Representations , 2015.
23	LIU Z, LIN Y T, CAO Y, et al. Swin transformer: hierarchical vision transformer using shifted windows[C]// Proc. of the IEEE/CVF International Conference on Computer Vision, 2021: 10012−10022.
24	VARGHESE R, SAMBATH M. YOLOv8: a novel object detection algorithm with enhanced performance and robustness[C]// Proc. of the International Conference on Advances in Data Engineering and Intelligent Computing Systems, 2024.
25	LIU J F, LIU Z B. YOLOv5s-BC: an improved YOLOv5s-based method for real-time apple detection[J]. Journal of Real-Time Image Processing, 2024, 21 (3): 88. doi: 10.1007/s11554-024-01473-1
26	扈琪, 胡绍海, 刘帅奇. 基于多层显著性模型的SAR图像舰船目标检测[J]. 系统工程与电子技术, 2024, 46 (2): 478- 487.
	HU Q, HU S H, LIU S Q. Ship detection in SAR image based on multi-layer saliency model[J]. Systems Engineering and Electronics, 2024, 46 (2): 478- 487.
27	YE H C, ZHANG B, CHEN T, et al. Performance-aware approximation of global channel pruning for multitask CNNs[J]. IEEE Trans. on Pattern Analysis and Machine Intelligence, 2023, 45 (8): 10267- 10284. doi: 10.1109/TPAMI.2023.3260903
28	YU Y, WANG C P, FU Q, et al. Techniques and challenges of image segmentation: a review[J]. Electronics., 2023, 12, 1199. doi: 10.3390/electronics12051199
29	于营, 王春平, 付强, 等. 语义分割评价指标和评价方法综述[J]. 计算机工程与应用, 2023, 59 (6): 57- 69. doi: 10.3778/j.issn.1002-8331.2207-0139
	YU Y, WANG C P, FU Q, et al. Survey of evaluation metrics and methods for semantic segmentation[J]. Computer Engineering and Applications, 2023, 59 (6): 57- 69. doi: 10.3778/j.issn.1002-8331.2207-0139
30	CAI Z W, VASCONCELOS N. Cascade R-CNN: delving into high quality object detection[C]// Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2018: 6154−6162.
31	REN S Q, HE K M, GIRSHICK R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Trans. on Pattern Analysis and Machine Intelligence, 2017, 39 (6): 1137- 1149. doi: 10.1109/TPAMI.2016.2577031
32	ZHENG Z H, YE R G, WANG P, et al. Localization distillation for dense object detection[C]// Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022: 9407−9416.
33	WANG C Y, BOCHKOVSKIY A, LIAO H Y M. YOLOv7: trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]// Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023: 7464−7475.
34	WANG C Y, YEH I H, LIAO H Y M. YOLOv9: learning what you want to learn using programmable gradient information[C]// Proc. of the 18th European Conference on Computer Vision, 2025.
35	WANG A, CHEN H, LIU L H, et al. YOLOv10: real-time end-to-end object detection [C]// Proc. of the 38th Conference on Neural Information Processing Systems, 2024, 37: 107984−108011.

阶段	层（类型）	Kernel 配置	输出尺寸
阶段1	Input	—	（640,640,3）
	Conv_1	4×4，3	（160,160,3）
	B_N_1	1×1，64	（160,160,64）
阶段2	Max_pooling	2×2	（80,80,64）
	Conv_2	3×3，64	（80,80,64）
	B_N_2	1×1，128	（80,80,128）
阶段3~阶段4	Max_pooling	2×2	（40,40,128）
	Conv_3/4	3×3，128	（40,40,128）
	B_N_3/4	1×1，256	（40,40,256）
阶段5~阶段7	Max_pool	2×2	（40,40,256）
	Conv_5/6/7	3×3，256	（40,40,256）
	B_N_5/6/7	1×1，512	（40,40,512）
阶段8~阶段9	Max_pooling	2×2	（40,40,512）
	Conv_8/9	3×3，512	（40,40,512）
	B_N_8/9	1×1，1024	（40,40,1024）
阶段10~阶段11	Max_pooling	2×2	（20,20,1024）
	Conv_10/11	3×3，1024	（20,20,1024）
	B_N_10/11	1×1，1024	（20,20,1024）

名称	配置信息
操作系统	Ubuntu20.04
CPU	Intel（R） Core（TM） i9-9900K CPU @ 3.60GHz
显卡GPU	NVIDIA RTX4090, 24GB
CUDA版本	11.6
Torch	1.12.0+cu113/1.8.0+cu111
Torch_audio	0.12.0+cu113/0.8.0
Torch_vision	0.13.0+cu113/0.9.0+cu111

模型	GFLOPs	参数/×10⁶	mAP^test@50	mAP^test@[0.5:0.95]
Cascade-RCNN ^[31]	301.0	88.015	0.663	0.392
Faster-RCNN ^[32]	419.0	99.246	0.652	0.352
定位蒸馏^[33]	267.0	51.251	0.431	0.193
YOLOv5s ^[25]	108.2	46.144	0.695	0.417
YOLOv7 ^[26]	103.2	35.632	0.623	0.418
YOLOv8l ^[24]	164.8	43.608	0.686	0.423
YOLOv9c ^[34]	236.6	48.352	0.768	0.426
YOLOv10b ^[35]	97.7	19.936	0.693	0.344
YOLOv10l ^[35]	126.5	25.117	0.698	0.364
本文算法	111.2	40.691	0.729	0.436

GFLOPs	参数/×10⁶	mAP^test@0.5	mAP^test@[0.5:0.95]	YOLOv5s主干网络替换
GFLOPs	参数/×10⁶	mAP^test@0.5	mAP^test@[0.5:0.95]	VanillaNet_5	VanillaNet_10	VanillaNet_13	SwinTransformer
93.5	23.078	0.617	0.339（↓9.36%）	√	—	—	—
160.6	53.249	0.652	0.387（↑3.48%）	—	√	—	—
200.9	66.145	0.695	0.418（↑11.76%）	—	—	√	—
75.6	29.107	0.723	0.394（↑5.35%）	—	—	—	√

GFLOPs	参数/×10⁶	mAP^test@0.5	mAP^test@[0.5:0.95]	YOLOv8l主干网络替换
GFLOPs	参数/×10⁶	mAP^test@0.5	mAP^test@[0.5:0.95]	VanillaNet_5	VanillaNet_7	VanillaNet_10	VanillaNet_13
96.7	23.957	0.686	0.395（↑7.05%）	√	—	—	—
123.5	41.232	0.698	0.417（↑13.01%）	—	√	—	—
163.8	54.128	0.726	0.415（↑12.47%）	—	—	√	—
204.1	67.024	0.686	0.441（↑19.51%）	—	—	—	√

PE-Net: a real-time landslide detection network with optimized pruning

RichHTML

PDF (PC)

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 20

References 35

Related Articles 15

Recommended Articles

Metrics

Comments

模型	检测头	GFLOPs	参数/×10⁶	mAP^test@0.5	mAP^test @[0.5:0.95]
YOLOv8l	—	28.6	11.167	0.641	0.361
VanillaNet 10+C3	Dyhead 4	282.6	88.788	0.697	0.399（↑10.53%）
VanillaNet 13 +C3	Dyhead 4	363.4	114.304	0.724	0.428（↑18.56%）

模型	动态检测头	GFLOPs	参数/×10⁶	mAP^test@0.5	mAP^test@0.5:0.95
VanillaNet13+C3	Dyhead 2	363.4	114.230	0.723	0.418 （↑15.79%）
VanillaNet13+C3	Dyhead 4	363.4	114.304	0.724	0.428 （↑18.56%）
VanillaNet13+C3	Dyhead 6	363.4	114.378	0.685	0.424 （↑17.45%）

[1]	Jie JIANG, Limin ZHANG, Kai LIU, Wenjun YAN, Meng WANG. Infrared ship target detection method based on task alignment learning [J]. Systems Engineering and Electronics, 2025, 47(1): 34-40.
[2]	Zishuo HAN, Xiquan FAN, Qiang FU, Chuanyan MA, Dongdong ZHANG. Target detection based on multi-source information fusion from the perspective of drones [J]. Systems Engineering and Electronics, 2025, 47(1): 52-61.
[3]	Meng WANG, Bing ZHU. Application of uncertainty modeling in 2D and 3D object detection [J]. Systems Engineering and Electronics, 2023, 45(8): 2370-2376.
[4]	Daozhi WEI, Zhaoyu ZHANG, Jiahao XIE, Ning LI. Multi-sensor cross cueing technique based on improved Actor-Critic algorithm [J]. Systems Engineering and Electronics, 2023, 45(6): 1624-1632.
[5]	Renfei CHEN, Yong PENG, Zhongwen LI. A novel detector for floating objects based on continual unsupervised domain adaptation strategy [J]. Systems Engineering and Electronics, 2023, 45(11): 3391-3401.
[6]	Shuang SONG, Yue ZHANG, Linna ZHANG, Yigang CEN, Yidong LI. Lightweight target detection algorithm based on deep learning [J]. Systems Engineering and Electronics, 2022, 44(9): 2716-2725.
[7]	Hongyao LI, Xiaoqiang LI, Xinzhong HAN, Xueli XIE, Jianxiang XI. Cooperative object detection and recognition algorithm for multiple UAVs based on decision fusion [J]. Systems Engineering and Electronics, 2022, 44(3): 746-754.
[8]	Jiali FAN, Shaobing TIAN, Kui HUANG, Xingdong ZHU. Multi-scale object detection algorithm for aircraft carrier surface based on Faster R-CNN [J]. Systems Engineering and Electronics, 2022, 44(1): 40-46.
[9]	Yiming ZHANG, Jianliang AI. Positioning of aerial refueling drogue and docking control based on binocular vision [J]. Systems Engineering and Electronics, 2021, 43(10): 2940-2953.
[10]	Zhenzhen YANG, Jun LE, Yongpeng YANG, Lu FAN. Object detection algorithm of nonconvex motion-assisted low rank and sparse decomposition [J]. Systems Engineering and Electronics, 2020, 42(6): 1218-1225.
[11]	Juan SU, Long YANG, Hua HUANG, Guodong JIN. Improved SSD algorithm for small-sized SAR ship detection [J]. Systems Engineering and Electronics, 2020, 42(5): 1026-1034.
[12]	ZHOU Long, WEI Suyuan, CUI Zhongma, FANG Jiaqi, YANG Xiaoting, YANG Long. Multiobjective detection of complex background radar imagebased on deep learning [J]. Systems Engineering and Electronics, 2019, 41(6): 1258-1264.
[13]	GUO Shao-jun, SHEN Tong-sheng, XU Jian, MA Xin-xing. Detection of multi-ship targets at sea based on ObjectNess BING [J]. Systems Engineering and Electronics, 2016, 38(1): 14-20.
[14]	ZHOU Jun-jing, DUAN Jian-min. Moving object detection for intelligent vehicles based on occupancy grid map [J]. Systems Engineering and Electronics, 2015, 37(2): 436-442.
[15]	DI Xiao-guang, FENG Xiao-dan, Fu Shao-wen. Moving object detection in non-planar scenes based on improved gradient suppression [J]. Systems Engineering and Electronics, 2014, 36(6): 1021-1026.

模型	GFLOPs	参数/×10⁶	mAP^test@50	mAP^test@[0.5:0.95]	FPS
不剪枝	363.4	114.429	0.724	0.428	31.94
Pruned 1	141.0	48.242	0.665 （↓8.15%）	0.396 （↓7.48%）	68.49 （↑114.43%）
Pruned 2	141.1	49.103	0.684 （↓5.52%）	0.414 （↓3.27%）	66.23 （↑107.36%）
Pruned 3	126.3	46.526	0.674 （↓6.91%）	0.395 （↓7.71%）	75.19 （↑135.40%）
Pruned 4	127.9	43.108	0.672 （↓7.18%）	0.410 （↓4.21%）	72.46 （↑126.87%）
Pruned 5	111.2	40.691	0.729 （↓0.55%）	0.436 （↑1.87%）	75.19 （↑135.41%）
Pruned 6	60.9	24.483	0.687 （↓5.11%）	0.404 （↓5.61%）	109.89 （↑244.05%）
Pruned 7	27.6	16.542	0.690 （↓4.70%）	0.374 （↓12.62%）	175.44 （↑449.28%）
Pruned 8	21.7	13.611	0.575 （↓20.58%）	0.317 （↓25.93%）	196.08 （↑513.90%）
Pruned 9	19.2	13.874	0.397 （↓45.17%）	0.221 （↓48.36%）	208.33 （↑552.26%）