Abstract
This paper proposes a lightweight moving object prediction system to detect and recognize pedestrian crossings, vehicles cutting-in, and vehicles ahead applying emergency brakes based on a 3D Convolution network for behavior prediction. The proposed design significantly improves the performance of the conventional 3D convolution network (C3D) adapted to predict the behaviors employing behavior recognition network capable of performing object localization, which is pivotal in detecting the numerous moving objects’ behaviors, combining and verifying the detected objects with the results of the YOLO v3 detection model with that of the proposed C3D model. Since the proposed system is a lightweight CNN model requiring far lesser parameters, it can be efficiently realized on an embedded system for real-time applications. The proposed lightweight C3D model achieves 10 frames per second (FPS) on a NVIDIA Jetson AGX Xavier and yields over 92.8% accuracy in recognizing pedestrian crossing, over 94.3% accuracy in detecting vehicle cutting-in behavior, and over 95% accuracy for vehicles applying emergency brakes.
Highlights
With the breakthrough of Artificial Intelligence (AI), technology is swiftly progressing—from traditional machine learning methods to deep learning methods that uses neural networks (NNs) and from image classification networks to object detection networks.the technological evolutions of both hardware and software designs have enabled AI networks to own the capability to judge just like human beings, or sometimes, even better than human beings.Advanced Driving Assistance Systems (ADAS) are electronic systems embedded as a crucial part of modern day motor vehicles, increasing not just the user-friendliness of the motor vehicles and enhancing the overall safety of the passengers as well as the pedestrians
This paper proposes a lightweight 3D convolution model with heatmap layers using fully-connected layers and unpooling layers to detect pedestrians crossing, vehicles cutting-in, and vehicles applying emergency brakes
The proposed design is optimized for computational low complexity and to possess a smaller model size, as low as 5.7 MB, to be convenient enough to realize in embedded systems like NVIDIA Jetson Xavier, which most state-of-the-art methods lack in respect to multiple object detection in real-time for ADAS applications
Summary
With the breakthrough of Artificial Intelligence (AI), technology is swiftly progressing—from traditional machine learning methods to deep learning methods that uses neural networks (NNs) and from image classification networks to object detection networks.the technological evolutions of both hardware and software designs have enabled AI networks to own the capability to judge just like human beings, or sometimes, even better than human beings.Advanced Driving Assistance Systems (ADAS) are electronic systems embedded as a crucial part of modern day motor vehicles, increasing not just the user-friendliness of the motor vehicles and enhancing the overall safety of the passengers as well as the pedestrians. A group of applications such as the Electronics 2021, 10, 692 it is desirable to have a highly accurate ADAS system to assess any situation in real-time while driving and alert the driver, and at times take certain decisions. Other additional efforts are necessary to obtain and maintain the temporal information On hand, behavior recognition based on convolution can get temporal information the other hand, behavior recognition based on 3D convolution can get temporal by just performing the 3D convolution always suffers from a large number information by justconvolution performingbut convolution but the 3D convolution always suffers from parameters requires supplemental efforts to fit a 2Defforts pre-trained into a 3D aoflarge numberand of parameters and requires supplemental to fit amodel model into a 3D network structure. The process of behavior recognition based on 2D convolution solely extracts spatial features and it needs other ways to obtain temporal information
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
