To address the problems of low resolution and blurred details in highway images caused by factors such as rain and fog, illumination interference, and nighttime lighting, this paper proposes a highway traffic image enhancement algorithm based on improved GAN in complex weather conditions. The attention mechanism and the multiscale feature fusion were combined to improve the generator network, which could effectively reduce noise while improving the attention of high-frequency region information. The improved PatchGAN in the discriminator used a local discrimination strategy to distinguish the generated image from the real image, and then the Nash equilibrium was achieved through the continuous interaction between the generator and the discriminator, to ensure the integrity and authenticity of the restored image. Compared with other image enhancement algorithms, using PSNR and SSIM as measurement indicators, the experimental results showed that the proposed algorithm’s results were, respectively, 21.97% and 12.89% higher in nighttime enhancement, 26.16% and 12.75% higher in rain removal, and 26.56% and 12.1% higher in fog removal. The proposed algorithm can not only retain the image details and feature information, but also produce effective denoising, which increases the reliability of image-based traffic information processing and analysis.

Effectively extracting features from a discrete point cloud is necessary for three-dimensional (3D) vehicle tracking. However, a point cloud data set is large and sparsely distributed, hindering the success of vehicle-tracking algorithms. To solve this problem, this paper proposes a 3D vehicle object tracking algorithm based on bounding box similarity measurement. The algorithm includes state prediction, temporal association, trajectory management, state update, and other processes. Also incorporated is a vehicle object temporal association method based on a siamese encoder. The bounding box is encoded into a high-dimensional space, the feature distance is calculated as the time series association cost, and triplet loss was introduced to urge the encoder to learn the geometric similarity of the truth matching box. A 3D Kalman filter and greedy matching are used to effectuate 3D vehicle object tracking algorithm. The experimental results using a KITTI multi-object tracking dataset show that the proposed algorithm can achieve good vehicle tracking performance. In the case of frame loss of point cloud in the frequency reduction simulation, compared with the benchmark method AB3DMOT, the average multi-object tracking accuracy ( AMOTA) and the average multi-object tracking accuracy ( AMOTP) of the improved method are increased by 2.62% and 1.41% respectively, and the performance is better in the frame loss scene.

As one of the early manifestations of road pavement structure degradation, road cracks will accelerate the deterioration of the road if not detected and repaired in time. Aiming at the problems of low recall and incomplete crack detection in current road detection, based on the U-Net network, this paper proposed an Attention-Dynamic Snake Convolution U-Net (ADSC-U-Net) network. Firstly, the dynamic snake-shaped convolution was added to the normal downsampling process to make the network adaptively focus on the slender and curved local features, which can solve the problem of low accuracy of small crack detection. Secondly, the attention mechanism was used to pay better attention to the significant features of positive samples under the condition of a large proportion gap between positive and negative samples, which solved the problem of the poor crack integrity detection effect. Finally, the dataset was expanded by random vertical and horizontal flip operations, which solved the problem of network training overfitting caused by the small-scale datasets. The experimental results showed that, when the input image had a resolution of 480 × 320, evaluation indices P, R, and F1 of ADSC-U-Net on the self-built dataset were 74.44%, 68.77%, and 69.42%, respectively. Compared to SegNet, DeepLab, and DeepCrack, the P was improved by 1.90%, 2.49%, and 11.64%, respectively; the R was improved by 8.01%, 4.70%, and 59.58%, respectively; and the comprehensive evaluation index F1 was improved by 5.73%, 4.02%, and 55.87%, respectively, which proves the effectiveness of the proposed method.

Short-term traffic flow has the characteristics of complex, changeable, strong timeliness, and so on. So the traditional prediction algorithm is difficult to meet its high real-time and accuracy requirements. In this paper, a multiscale and high-precision LSTM-GASVR short-term traffic flow prediction algorithm is proposed. This method uses 15 min traffic flow data of the first 16 sections as input and completes the data preprocessing operation through reconstruction, normalization, and rising dimension by working day factor; establishing the prediction model based on the long- and short-term memory network (LSTM) and inverse normalization; and proposing the GA-SVR model to optimize the prediction results, so as to realize the real-time high-precision prediction of traffic flow. The prediction experiment is carried out according to the charge data of a toll station in Xi’an, Shaanxi Province, from May 2018 to May 2019. The comparison and analysis of various algorithms show that the prediction algorithm proposed in this paper is 20% higher than the LSTM, GRU, CNN, SAE, ARIMA, and SVR, and the R² can reach 0.982, the explanatory variance is 0.982, and the MAPE is 0.118. The proposed traffic flow prediction algorithm provides strong support for traffic managers to judge the state of the road network to control traffic and guide traffic flow.

In vehicle visual navigation, image matching algorithm is highly critical to positioning accuracy and processing efficiency. One single matching algorithm cannot satisfy all types of image features accurate acquisition, so Harris, SUSAN, FAST, SIFT, and SURF are respectively adopted to process various road images under normal lighting condition. During practical application, the appropriate algorithm can be selected based on detection rate and running time of the above algorithms. Aiming at the illumination change interference of the collected images in vehicle visual navigation, many traditional matching algorithms for illumination change are not optimal, so an image precise matching algorithm with illumination change robustness is proposed. Because image edges and detail information have lower sensitivity for illumination change, SURF feature points are optimized by image gradient based on the idea of Canny, and the bidirectional search is used to obtain precise matching points. The experimental results show that feature point detection of the algorithm remains good stability for illumination change in images, and the matching accuracy can reach more than 94%. The algorithm is not only robust to illumination change, but also ensures higher matching speed and meanwhile improves the matching accuracy significantly.

For face recognition systems, liveness detection can effectively avoid illegal fraud and improve the safety of face recognition systems. Common face attacks include photo printing and video replay attacks. This paper studied the differences between photos, videos, and real faces in static texture and motion information and proposed a living detection structure based on feature fusion and attention mechanism, Dynamic and Texture Fusion Attention Network (DTFA-Net). We proposed a dynamic information fusion structure of an interchannel attention block to fuse the magnitude and direction of optical flow to extract facial motion features. In addition, for the face detection failure of HOG algorithm under complex illumination, we proposed an improved Gamma image preprocessing algorithm, which effectively improved the face detection ability. We conducted experiments on the CASIA-MFSD and Replay Attack Databases. According to experiments, the DTFA-Net proposed in this paper achieved 6.9% EER on CASIA and 2.2% HTER on Replay Attack that was comparable to other methods.