In the field of autonomous driving, ensuring safe switching of vehicle control is the key to driving safety. Currently, most autonomous driving systems on the market are still in the stage of human-machine collaborati...
详细信息
ISBN:
(纸本)9798400712739
In the field of autonomous driving, ensuring safe switching of vehicle control is the key to driving safety. Currently, most autonomous driving systems on the market are still in the stage of human-machine collaborative driving, and often need to switch between people and systems. Therefore, it is particularly important to accurately identify the driver's grip status in realtime, which will directly affect the safe driving of the system. This paper introduces an algorithm that uses image recognition technology to detect whether the driver's hands are off the steering wheel. The Bezier curve fitting technology is mainly used to extract the steering wheel image, and the EMD algorithm is used to compare the images to determine whether the driver has abnormal behavior. The accuracy of the algorithm is evaluated by selecting data sets under different lighting conditions. Through the evaluation of the three indicators of accuracy, false detection rate, and missed detection rate, the accuracy of the algorithm in actual driving exceeds 91%, and the average processingtime is 67.4 ms, which meets the real-time requirements. Compared with the method based on skin color extraction, the algorithm is more robust to lighting changes. The results show that the algorithm can effectively improve driving safety and has application potential in the field of intelligent driving in the future, which can provide a basis for future research on driving behavior detection based on deep learning.
The proceedings contain 33 papers. The topics discussed include: measuring environment color reflection for integrated CGI in live-action footage;target tracking based on multiple appearance model fusion with sparse r...
ISBN:
(纸本)9781450376648
The proceedings contain 33 papers. The topics discussed include: measuring environment color reflection for integrated CGI in live-action footage;target tracking based on multiple appearance model fusion with sparse representation;research of express box defect detection based on machine vision;defect detection in id cards with accurately reconstructed reference image;flying point target tracking using infrared images;two-stage metric learning for cross-modality person re-identification;pavement type recognition based on deep learning;real-time head pose estimation based on face geometry;and image recognition based on multi-scale dilated lightweight network model.
We study the problem of online learning of optimal offloading policies for imageprocessing tasks, for minimizing a cost that is weighted sum of transmit energy and object recognition error rate. A mobile node generat...
详细信息
ISBN:
(纸本)9781538674628
We study the problem of online learning of optimal offloading policies for imageprocessing tasks, for minimizing a cost that is weighted sum of transmit energy and object recognition error rate. A mobile node generates imageprocessing tasks that involve object recognition. There exist three options: (i) transmit the image to a remote server for processing with a deep-learning (DL) model, (ii) process locally with a simpler model, (iii) apply a lightweight, error-prone technique for object detection, and if objects are detected, then send image to the server. The proper offloading decision requires knowledge of the transmit energy cost and object recognition error rate for each option. However, these processes are non-stationary due to unpredictable object occurrence, mobility and propagation dynamics, and they depend on the object inference result which is unknown at decision time. We cast the problem as an adversarial multi-armed bandit, in which the EXP3 algorithm achieves sublinear regret. For the constrained problem, we propose an algorithm that extends EXP3 and achieves good regret in the objective and constraint, thus asymptotically learning the optimal static randomized offloading policy, while satisfying the error constraint. Performance is validated via numerical experiments informed by real-life object recognition measurements and models.
The predominant function of most facial analysis systems revolves around facial alignment and eye tracking, crucial for locating key facial landmarks in images or videos. While developers have access to various models...
详细信息
Aiming at the problem that it is difficult to detect thread surface defects, a set of thread surface defect detection system based on imageprocessing is developed. It adopts non-contact, high-precision and real-time ...
详细信息
The proceedings contain 60 papers. The topics discussed include: a light weight CNN algorithm for the detection of various tomato plant diseases;a novel convolutional neural network classifier for real-time traffic si...
ISBN:
(纸本)9798350375466
The proceedings contain 60 papers. The topics discussed include: a light weight CNN algorithm for the detection of various tomato plant diseases;a novel convolutional neural network classifier for real-time traffic sign detection;brain controlled robotic car using mind wave;crop sentry: ai-enhanced protection for agriculture field;deep learning-based modular framework for risk mitigation through crowd counting;FPGA implementation of exudates detection in fundus images through machine learning;license plate recognition using federated learning;machine learning driven exploration and identification of steel surfaces: leveraging advanced computer vision techniques;and unified image mastery : an integrated approach to image enhancement, object detection and classification.
Dehazing algorithms have been developed in re-sponse to the need for effectively and instantaneously removing atmospheric turbidities such as mist, haze, and fog from media. The removal of haze from an image or video ...
详细信息
With the rapid improvement of UAV imaging equipment in terms of image resolution and video frame rate, relying solely on software to enhance images with extremely high data volume is no longer sufficient to meet real-...
详细信息
Laser additive manufacturing (LAM) demands robust real-time monitoring to identify defects and ensure product quality. Traditional methods, mainly reliant on coaxial cameras, lacks placement flexibility. Further compl...
详细信息
ISBN:
(纸本)9798350373172;9798350373189
Laser additive manufacturing (LAM) demands robust real-time monitoring to identify defects and ensure product quality. Traditional methods, mainly reliant on coaxial cameras, lacks placement flexibility. Further complexities arise from the inherent variability in melt pool geometries and the high computational demands of imageprocessing, hindering effective real-time monitoring. This research proposes a novel technique to directly infer melt pool visual characteristics in LAM by synergizing acoustic signal features with robotic tool-center-point (TCP) motion data. Acoustic monitoring has shown great promise in tasks typically reliant on vision sensors. In addition, the dynamics of the LAM process and defect occurrences are spatially dependent, primarily due to heat accumulation. By combining acoustic signals with spatial data from robot TCP motion, our method tracks melt pool variations with an R-2 score above 0.7. An ablation study demonstrated that the proposed method outperforms the acoustic-only models. The findings suggest that the integration of a simple microphone sensor with robot motion information emerges as a flexible, cost-effective alternative for capturing dynamic melt pool behavior. It presents new prospects for closed-loop control in the LAM process.
This paper focuses on traditional deep learning-based no-reference (or reference-based) image quality assessment(IQA) methods, enhancing them from the perspective of image feature extraction. It replaces the VGG16 net...
详细信息
暂无评论