Search

Removing physical scanners on mobile devices has come to be one of the most recent trends, adopting in-display scanners becomes the solution of mobile device authorization applications. As the characteristics of the image captured from in-display scanners were quite different from physical scanners, the criteria of the image quality should be reconsidered to maintain the effectiveness. This paper proposes an image quality analysis technique for optical-based fingerprint on display (FoD) applications, we evaluated the method on our own dataset with 11520 images, and proof that it is truly useful to fingerprint images captured from optical-based in-display scanners.

License plate detection has an important role to play in intelligent transportation systems. It aims at locating the license plate and accurate region information in the image. In recent years, there are lots of efficient detection methods based on the application of deep convolutional networks proposed. Although these approaches have presented promising results, these methods use horizontal or nearly horizontal bounding boxes to locate the object which is inappropriate for multi- oriented object in the real world. In this paper, a new license plate detection framework has been proposed for oriented license plate detection which combines the oriented region with a deep learning method to recognize oriented license plate detection. The proposed framework contains three sub- networks for extracting, generating and prediction proposals. Experimental results for license plate detection from the Chinese City Parking Dataset demonstrate that the proposed approach has good performance under specific scenarios.

Numerous vehicle detection methods have been proposed to obtain trustworthy traffic data for the development of intelligent traffic systems. Most of these methods perform sufficiently well under common scenarios, such as sunny or cloudy days, however, the detection accuracy drastically decreases under various bad weather conditions, such as rainy days or days with glare, which normally happens during sunset. This study proposes a vehicle detection system with a visibility complementation module that improves detection accuracy under various bad weather conditions. Furthermore, the proposed system can be implemented without retraining the deep learning models for object detection under different weather conditions. The complementation of the visibility was obtained through the use of a dark channel prior and a convolutional encoder&ndash, decoder deep learning network with dual residual blocks to resolve different effects from different bad weather conditions. We validated our system on multiple surveillance videos by detecting vehicles with the You Only Look Once (YOLOv3) deep learning model and demonstrated that the computational time of our system could reach 30 fps on average, moreover, the accuracy increased not only by nearly 5% under low-contrast scene conditions but also 50% under rainy scene conditions. The results of our demonstrations indicate that our approach is able to detect vehicles under various bad weather conditions without the need to retrain a new model.

Vehicle re-identification is a popular issue in intelligent traffic research. A lot of proposed method achieve vehicle re-identification by recognizing their license plate, because of the uniqueness. However, license plate can be stolen and pinned to different vehicles by criminals to hide their identities. In addition, the license plate number might be covered by dirt or stain, even hided from different viewpoints, makes the character recognition result might be wrong or unrecognized. To get more robust re-identification result, not only the license plate should be considered, but also the appearance. In this paper, we adopt Siamese Convolutional Neural Network structure, take license plate and vehicle appearance as input, come up with a neural network for vehicle re-identification task. We validate our proposed method on VeRi-776 dataset, and proof that it can deal with vehicle re-identification task well, even under variant viewpoints scenarios.

Foreseeing the positions of the surrounding objects is a critical technique to autonomous vehicles. Referring to the positions predicted by the trajectories, moving vehicles could avoid possible collisions. Many trajectory prediction methods about have been proposed, from traditional hand-crafted approaches to Recurrent Neural Network (RNN) models. Although the proposed methods can achieve satisfactory results on publicly available human-trajectory datasets, there are still exist some problems, such as predicting positions in obstacles, sharp turns that are not socially acceptable. We adopted the concepts of social GAN to propose a new model that take the social forces, scene information and motion direction into account when predicting trajectories. We demonstrate the performance of our proposed methods by adopting several videos from Stanford Drone Dataset.

This study proposes a vision-based nightime surrounding nighttime vehicle detection system implemented by automatic multilevel thresholding method, denoise, connected component labeling, head light matching method, and head light tracking and classification. It is able to detect head light position, and get the vehicle position. Testing by four videos that we have record while we were driving, for side view, accuracy are all higher than 90%, and the time cost are all higher than 140 frames per second; for rear view, accuracy are all higher than 90%, and the time cost are all higher than 85 frames per second. By this system, we can make sure that there is any vehicle surrounding or not. If there is, it will decide which kind of vehicle it is, and notice the driver with different level, to supply more information and longer reaction time for driver, in order to decrease accident rate, increase driving safety.

In this paper, Haar-like patterns and a classification machine were employed for fingerprint feature extraction. Although the Gabor filter has high accuracy for feature extraction under a wide angle representation range, the full range of the filter is not required. Moreover, fingerprints have clear direction field representation. Due to these two key factors, this paper uses modified Haar-like patterns to create near-circular patterns for acceptable feature detection. A support vector machine classifier is used to detect features. In a performance comparison that used 15 general Haar-like patterns as a benchmark, the proposed algorithm was able to reduce the average computation required by approximately 50% without sacrificing accuracy. Thus, the proposed method is suitable for real-world applications.

Most object detection models cannot achieve satisfactory performance under nighttime and other insufficient illumination conditions, which may be due to the collection of data sets and typical labeling conventions. Public data sets collected for object detection are usually photographed with sufficient ambient lighting. However, their labeling conventions typically focus on clear objects and ignore blurry and occluded objects. Consequently, the detection performance levels of traditional vehicle detection techniques are limited in nighttime environments without sufficient illumination. When objects occupy a small number of pixels and the existence of crucial features is infrequent, traditional convolutional neural networks (CNNs) may suffer from serious information loss due to the fixed number of convolutional operations. This study presents solutions for data collection and the labeling convention of nighttime data to handle various types of situations, including in-vehicle detection. Moreover, the study proposes a specifically optimized system based on the Faster region-based CNN model. The system has a processing speed of 16 frames per second for 500 &times, 375-pixel images, and it achieved a mean average precision (mAP) of 0.8497 in our validation segment involving urban nighttime and extremely inadequate lighting conditions. The experimental results demonstrated that our proposed methods can achieve high detection performance in various nighttime environments, such as urban nighttime conditions with insufficient illumination, and extremely dark conditions with nearly no lighting. The proposed system outperforms original methods that have an mAP value of approximately 0.2.