Multi-scale traffic vehicle detection based on faster R–CNN with NAS optimization and feature enrichment

Bai

Computational Intelligence and Neuroscience

et al. 2022

Traffic sign detection is a challenging problem in the field of unmanned driving, particularly important in complex environments. We propose a method, based on the improved You only look once (YOLO) v4, to detect and recognize multiscale traffic signs in complex environments. This method employs an image preprocessing module that can classify and denoize images of complex environments and then input the images into the improved YOLOv4. We also design an improved feature pyramid structure to replace the original feature pyramid of YOLOv4. This structure uses an adaptive feature fusion module and a multiscale feature transfer mechanism to reduce putative information loss in the feature map generation process and improve the information transfer between deep and shallow features, enhancing the representation ability of feature pyramids. Finally, we use EIOU LOSS and Cluster-NMS to further improve the model performance. The experimental results on the fusion of Tsinghua-Tencent 100 K and our collected dataset show that the proposed method achieves an mAP of 81.78%. Compared to existing methods, our method demonstrates its superiority with regard to traffic sign detection.

Section: Related Workmentioning

confidence: 99%

Multiscale Traffic Sign Detection Method in Complex Environment Based on YOLOv4

Bai

Computational Intelligence and Neuroscience

et al. 2022

Computational Intelligence and Neuroscience

“…Luo et al [ 53 ] presented a model based on the faster RCNN with NAS optimization and feature enrichment to perform the effective detection of multiscale vehicle targets in traffic scenes. Luo et al proposed a Retinex-based image adaptive correction algorithm (RIAC) (to reduce the influence of shadows and illumination), conducted neural architecture search (NAS) on the backbone network used for feature extraction of the faster RCNN (to generate the optimal cross-layer connection to extract multilayer features more effectively), and used the object feature enrichment that combines the multilayer feature information and the context information of the last layer after cross-layer connection (to enrich the information of vehicle targets and improve the robustness of the model for challenging targets such as small scale and severe occlusion).…”

Section: Related Workmentioning

confidence: 99%

Robust Real‐Time Traffic Surveillance with Deep Learning

Fernandez

Plaza

Fernández

et al. 2021

Real-time vehicle monitoring in highways, roads, and streets may provide useful data both for infrastructure planning and for traffic management in general. Even though it is a classic research area in computer vision, advances in neural networks for object detection and classification, especially in the last years, made this area even more appealing due to the effectiveness of these methods. This study presents TrafficSensor, a system that employs deep learning techniques for automatic vehicle tracking and classification on highways using a calibrated and fixed camera. A new traffic image dataset was created to train the models, which includes real traffic images in poor lightning or weather conditions and low-resolution images. The proposed system consists mainly of two modules, first one responsible of vehicle detection and classification and a second one for vehicle tracking. For the first module, several neural models were tested and objectively compared, and finally, the YOLOv3 and YOLOv4-based network trained on the new traffic dataset were selected. The second module combines a simple spatial association algorithm with a more sophisticated KLT (Kanade–Lucas–Tomasi) tracker to follow the vehicles on the road. Several experiments have been conducted on challenging traffic videos in order to validate the system with real data. Experimental results show that the proposed system is able to successfully detect, track, and classify vehicles traveling on a highway on real time.

“…The detection of vehicles, the most common form of transportation in remote sensing images, can provide data support for many fields. (2) For example, in the field of intelligent transportation system construction, vehicle detection can be applied to road condition information collection and road management information acquisition, and in the field of intelligent traffic management, vehicle detection can be applied to urban traffic flow statistics and road capacity control.…”

Section: Introductionmentioning

confidence: 99%

Vehicle Detection Algorithm Based on Background Features Assistance in Remote Sensing Image

Cao¹,

Bai²,

Pang³

et al. 2023

Sensors and Materials

Toward solving the problem of the lack of useful features caused by inconspicuous vehicle features and the interference of similar features around a vehicle in the process of remote sensing image vehicle detection, we propose an algorithm based on the assistance of background features. This algorithm is based on the YOLOv4 model and includes a weight redistribution module based on the feature correlation degree. The model introduces background features around a vehicle so that the model learns to determine the environment in which the vehicle target is located during training. This can effectively reduce the occurrence of missed detection. Moreover, the algorithm increases the sensitivity of the backbone to vehicle features through the weight redistribution of the features inside the anchor box, thus making full use of the feature correlation between the vehicle target and the surrounding background. As a result, the vehicle target is effectively distinguished from the interference target. The experimental results show that the precision rate and success rate of this algorithm on the DLR-3K dataset reach 75.4 and 68.5%, respectively. The model detection rate reached 12.7 frames/s. The proposed algorithm has high performance in executing vehicle detection tasks in the presence of interference due to similar targets.