Car Detection Research Articles

A Comparative Study of different Image Preprocessing Methods used for Multiple Car Detection and Tracking

A Comparative Study of different Image Preprocessing Methods used for Multiple Car Detection and Tracking

Automatic Detection and Classification of Road, Car, and Pedestrian Using Binocular Cameras in Traffic Scenes with a Common Framework

In order to solve the problems of traffic object detection, fuzzification, and simplification in real traffic environment, an automatic detection and classification algorithm for roads, vehicles, and pedestrians with multiple traffic objects under the same framework is proposed. We construct the final V view through a considerate U-V view method, which determines the location of the horizon and the initial contour of the road. Road detection results are obtained through error label reclassification, omitting point reassignment, and so an. We propose a peripheral envelope algorithm to determine sources of vehicles and pedestrians on the road. The initial segmentation results are determined by the regional growth of the source point through the minimum neighbor similarity algorithm. Vehicle detection results on the road are confirmed by combining disparity and color energy minimum algorithms with the object window aspect ratio threshold method. A method of multifeature fusion is presented to obtain the pedestrian target area, and the pedestrian detection results on the road are accurately segmented by combining the disparity neighbor similarity and the minimum energy algorithm. The algorithm is tested in three datasets of Enpeda, KITTI, and Daimler; then, the corresponding results prove the efficiency and accuracy of the proposed approach. Meanwhile, the real-time analysis of the algorithm is performed, and the average time efficiency is 13 pfs, which can realize the real-time performance of the detection process.

Car Plate Detection Based on Yolov3

An intelligent car plate detection method can make the travel more convenient and efficient. However, traditional methods are reasonably effective under the specific circumstances or strong assumptions only, and there are few databases for car plate detection. Therefore, a novel real-time car plate detection method based on improved Yolov3 has been proposed. In order to select the more precise number of candidate anchor boxed and aspect ratio dimensions, the K-Means algorithm is utilized. To solve the short of the available car plate database, a car plate database which has 6668 pictures has been established. As shown in the experimental results, the method which is proposed by this paper is better than original Yolov3. Thanks to the car plate database, the proposed method obtained better results even in the situation of inclination, too bright or too dark, different weather and so on.

Detection of the vehicle license plate using a kernel density with default search radius algorithm filter

With the contemporary improvements in intelligent transportation systems, automatic car License Plate Detection and Recognition (LPDR) has involved extensive research interests. It has a variety of potential applications in security and traffic control, and much work has been completed on the subject of LPDR. But, majority of prevailing procedures work well either at measured circumstances or with sophisticated image capture systems. It is still a challenging task to read license plates precisely in an open environment. The problem exists in extreme assortment in pattern of characters, such as dissimilar sizes, letterings, misrepresentation, obstruction or blurring, as well as high intricate backgrounds, similar to the common fonts in boards of shops, windows, handrails or construction blocks. In this study, a hybrid methodology has been established in recognizing license plates of candidates in a specific area. This research proposes a technique called kernel density function along with the binary technique that has been utilized for pre-processing. By means of the filtered binary value of the image, the position of car plate can be found by multiplying the binary value and original value of the image. The proposed method outperforms advanced techniques by a huge margin concerning about accuracy in detection as well as efficiency of run-time. The proposed method achieved 98.1 % accuracy to detect the License Plate (LP) and the computation cost is about 0.452 s.

A new car logo extraction method based on license plate detection

Nowadays, intelligent transportation system is becoming the new trend because of the development of modern traffic, which serves to help gather traffic data, control traffic and improve road safety. One big challenge within intelligent transportation system is car model detection, which provides important information for traffic monitoring and planning. Currently, most car model extraction focus more on one-pass solution, which directly finds out the car logo region. In this paper, a new hybrid car logo extraction method based on license plate detection is proposed. First, a text detection is applied on car images to locate the license plate. Then, within a region of interest above license plate, a sub region is extracted to locate the car logo based on edge information. The fundamental concept is that using the relative location information between the license plate and car logo can take the most advantage of existing state-of-the-art license plates detection methods which is practically accurate, robust and efficient. Experimental results on OpenALPR dataset demonstrate its good performance in extracting car logos.

Using Reinforcement Learning With Partial Vehicle Detection for Intelligent Traffic Signal Control

Intelligent Transportation Systems (ITS) have attracted the attention of researchers and the general public alike as a means to alleviate traffic congestion. Recently, the maturity of wireless technology has enabled a cost-efficient way to achieve ITS by detecting vehicles using Vehicle to Infrastructure (V2I) communications. Traditional ITS algorithms, in most cases, assume that every vehicle is observed, such as by a camera or a loop detector, but a V2I implementation would detect only those vehicles with wireless communications capability. We examine a family of transportation systems, which we will refer to as `Partially Detected Intelligent Transportation Systems'. An algorithm that can act well under a small detection rate is highly desirable due to gradual penetration rates of the underlying wireless technologies such as Dedicated Short Range Communications (DSRC) technology. Artificial Intelligence (AI) techniques for Reinforcement Learning (RL) are suitable tools for finding such an algorithm due to utilizing varied inputs and not requiring explicit analytic understanding or modeling of the underlying system dynamics. In this paper, we report a RL algorithm for partially observable ITS based on DSRC. The performance of this system is studied under different car flows, detection rates, and topologies of the road network. Our system is able to efficiently reduce the average waiting time of vehicles at an intersection, even with a low detection rate.

Probabilistic object detection and shape extraction in remote sensing data

Remote sensing mainly focuses on information extraction from data acquired by sensors on satellite and aerial platforms. Here, one such area of interest is ground object detection and shape extraction. Recently launched satellites and conventional aerial platforms (such as commercial UAV and professional drones) have sensors leading to more detailed and rich data source for this purpose. From these, data most of the times come in the form of optical images and LiDAR measurements. Resolution of this acquired data has increased significantly such that most ground objects (as buildings, trees, ships, cars, airplanes) can be detected and analyzed in detail. Therefore, computer vision methods have become extremely useful in remote sensing applications such as building detection and shape extraction for urban planning; tree crown measurement for crop yield forecasting; ship detection for monitoring unlawful fishery; car detection for traffic flow monitoring and intelligent transportation; and airplane detection for military and commercial operations. Researchers proposed several methods to automate the mentioned applications since manually handling them is extremely hard and prohibitively time consuming. Unfortunately, the proposed methods focus on one object type most of the times. Therefore, there is no general method to handle all the mentioned applications using computer vision tools. To overcome this problem, we propose a general framework for object detection and shape extraction in remote sensing data. Our method is based on probabilistic representation inspired by our previous work and perceptual organization principles. Due to space limitations, we only focus on buildings, trees, ships, airplanes, and cars as objects of interest in this study. We test the proposed method on several optical images acquired by different satellites and LiDAR data obtained from an aerial platform. For all objects of interest, we provide test results on both object detection and shape extraction steps. We analyze the proposed method based on these tests and discuss its strengths and weaknesses. We also comment on possible future extensions of the proposed method.

A Review Paper on License Plate Recognition System

ANPR provides solution in which the steps to run an efficient intelligent transport network might be taken. Owing to the rapid increase in vehicles it has become a requirement for traffic control management. ANPR's main goal is to track traffic and for the purpose of defense. Number plate recognition uses image processing techniques or OCR techniques and edge detection technology to detect characters on license plates. The model comprises of three modules which are module for car detection, module for license plate segmentation and module for recognition. Starting from auto robberies, violating traffic laws, to law enforcement administration, Image processing gave us a determination to put a stop to these violations. This review paper provided an examination of the different license plate recognition design implemented so far.

Density and Time based Traffic Control System using Video Processing

Traffic is the serious issue which each nation faces due to the expansion in number of vehicles. One of the strategies to beat the traffic issue is to build up a smart traffic control framework which depends on computing the traffic density and about utilizing constant video and picture preparing procedures. The topic is to control the traffic by deciding the traffic density on each roadside and control the traffic signal smartly by utilizing the density data. In this paper, an automated system based on processing of real time videos is proposed for detection of vehicles and recording count of them. The System will consist of various stages which includes Object Car Detection and Signal variation based on density. Captured video will be converted into frames and which will be pre-processed for object detection using Haar-Cascade than detected object count will be used to obtain the density and manipulate the signal accordingly. The density count algorithm works by contrasting the ongoing edge of live video by the reference picture and via looking through vehicles just in the district of intrigue (for example street region). The figured vehicle thickness can be contrasted and other course of the traffic so as to perform control of the traffic flags in more smart and proficient manner.

MIMO–OFDM communication systems for traffic data transmission in 5G drone small cells

A compressed sensing-based channel estimation (CE) method is proposed for doubly selective channels whilst multiple-input multiple-output (MIMO)–orthogonal frequency-division multiplexing (OFDM) system is employed. According to the simulation results, the proposed CE method, which is dubbed as the MIMO-linearising scheme performs better than the basic expansion model (BEM), which is a state-of-the-art CE method and applies frequency domain scattered pilots. Furthermore, MIMO-linearising can be utilised as an accurate procedure for estimating only the positions of non-zero channel taps in the scenarios where those positions change with a slower rate than amplitudes and phases of the channel taps. Based on that priori estimated positions (PEPs), modified versions of the linearising scheme and the BEM methods, called MIMO-linearising-PEP and BEM-PEP, respectively, are described, showing improvement on performance and spectral efficiency. In addition, by considering the physical characteristic of the vehicular channels, an alternative version of the MIMO-linearising-PEP, called MIMO-linearising-efficient, is described to reduce the computational complexity and increase the spectral efficiency further. Simulation results indicate that the proposed CE methods can measure the channel more accurately and enhance the car detection accuracy in the frames which have been used as the inputs of the communication systems.

SARPNET: Shape attention regional proposal network for liDAR-based 3D object detection

Real-time 3D object detection is a fundamental technique in numerous applications, such as autonomous driving, unmanned aerial vehicles (UAV) and robot vision. However, current LiDAR-based 3D object detection algorithms allocate inadequate attention to the inhomogeneity of LiDAR point clouds and the shape encoding capability of regional proposal schemes. This paper introduces a novel 3D object detection network called the Shape Attention Regional Proposal Network (SARPNET), which deploys a new low-level feature encoder to remedy the sparsity and inhomogeneity of LiDAR point clouds with an even sample method, and embodies a shape attention mechanism that learns the statistic 3D shape priors of objects and uses them to spatially enhance semantic embeddings. Experimental results show that the proposed one-stage method outperforms state-of-the-art one-stage and even two-stage methods on the KITTI 3D object detection benchmark. It achieved a BEV AP of (87.26%, 62.80%), 3D AP of (75.64%, 60.43%), and orientation AP of (88.86%, 71.01%) for the detection of cars and cyclists, respectively. Besides, the method is the third winner in the nuScenes 3D Detection challenge of CVPR2019 Workshop on Autonomous Driving (WAD).

IoT based car accident detection and notification algorithm for general road accidents

<p>With an increase in population, there is an increase in the number of accidents that happen every minute. These road accidents are unpredictable. There are situations where most of the accidents could not be reported properly to nearby ambulances on time. In most of the cases, there is the unavailability of emergency services which lack in providing the first aid and timely service which can lead to loss of life by some minutes. Hence, there is a need to develop a system that caters to all these problems and can effectively function to overcome the delay time caused by the medical vehicles. The purpose of this paper is to introduce a framework using IoT, which helps in detecting car accidents and notifying them immediately. This can be achieved by integrating smart sensors with a microcontroller within the car that can trigger at the time of an accident. The other modules like GPS and GSM are integrated with the system to obtain the location coordinates of the accidents and sending it to registered numbers and nearby ambulance to notify them about the accident to obtain immediate help at the location.</p>

Deep Learning RetinaNet based Car Detection for Smart Transportation Network

ABSTRAKDeteksi objek yang merupakan salah satu bagian utama dari sistem Smart Transportasion Network (STN) diajukan pada penelitian ini. Penelitian ini menggunakan salah satu model STN yaitu Infrastructure-to-Vehicle (I2V), dimana sistem ini bekerja dengan mendeteksi kendaraan mobil menggunakan model arsitektur RetinaNet dengan backbone Resnet101 dan FPN (Feature Pyramid Network), kemudian hasil deteksi mentrigger VLC transmitter yang terpasang di lampu penerangan jalan mengirimkan sinyal informasi menuju VLC receiver yang dipasang di mobil. Pada tahap proses training, jumlah dataset mobil yang digunakan adalah sekitar 1600 image dan 400 validation image serta pengulangan proses sebanyak 100 epoch. Berdasarkan 50 kali pengujian pada image test, diperoleh nilai precision mencapai 86%, nilai recall mencapai 85% dan f1-score mencapai 84%.Kata kunci: Object detection, RetinaNet, Resnet101, STN, VLC, I2V ABSTRACTObject detection is one of the main part in Smart Transportation Network (STN) system proposed in this research. This research used one of the STN models, namely Infrastructure-to-Vehicle (I2V), a system works by detecting car using RetinaNet architecture model with ResNet 101 and FPN (Feature Pyramid Network) as backbone, then the detection result triggers VLC transmitter set up on the street lighting to transmit information signal to the VLC receiver which set up in the car. At the training process stage, the number of car datasets is approximately 1600 images, 400 validation images and repetition of processes about 100 epochs. Based on the 50 times testing process on a image test, it is obtained 86% of a precision value, by reaching 85% of recall value, and 84% of f1-score. Keywords: Object detection, RetinaNet, Resnet101, STN, VLC, I2V

3D-GIoU: 3D Generalized Intersection over Union for Object Detection in Point Cloud

Three-dimensional (3D) object detection is an important research in 3D computer vision with significant applications in many fields, such as automatic driving, robotics, and human–computer interaction. However, the low precision is an urgent problem in the field of 3D object detection. To solve it, we present a framework for 3D object detection in point cloud. To be specific, a designed Backbone Network is used to make fusion of low-level features and high-level features, which makes full use of various information advantages. Moreover, the two-dimensional (2D) Generalized Intersection over Union is extended to 3D use as part of the loss function in our framework. Empirical experiments of Car, Cyclist, and Pedestrian detection have been conducted respectively on the KITTI benchmark. Experimental results with average precision (AP) have shown the effectiveness of the proposed network.

Person detection in natural scenes: A comparison with car detection.

Person detection in natural scenes: A comparison with car detection.

Automated Car Number Plate Recognition System using K-Nearest Neighbor

The Automatic number plate recognition (ANPR) is a mass reconnaissance strategy that utilizations optical character recognition on images to peruse the license plates on vehicles. The car number plate detection has the various phases like pre-processing, segmentation and classification. In the previous work, the morphological operation is applied for the car number plate detection. The morphological operation has the low accuracy for the car number plate detection. In the proposed work, the region based segmentation and K-nearest neighbor classification is applied for the character recognition. The proposed is implemented in MATLAB and results are analyzed in terms of accuracy.

Wheel flat detectability for Y25 railway freight wagon using vehicle component acceleration signals

A wheel flat is a common fault in railway vehicles that may develop after incorrect brake application, or localised wheel-rail adhesion reduction. Operators detect wheel flats using wayside systems that are not capable of delivering real-time alarms. A wheel flat defect was modelled and introduced to a railway wagon multibody dynamic simulation to study the vehicle dynamic behaviour triggered by such wheel tread defect. Bogie-based and car body-based wheel flat detection was evaluated to determine eventual hardware robustness and signal processing requirements for on-board wheel flat detection devices. Various vehicle operating conditionds and wheel flat defects were simulated to analyse the wheel flat impact propagation throughout the railway vehicle bodies. Wheel defect detectability was assessed using detectability indicators (DIs) based on envelope power spectrum analysis, crest factor and RMS. It was found that bogie and car body-based wheel flat detectability is highly dependent on the track characteristics. The wheel flat signature was reduced in the wheelset-bogie and bogie-car body interfaces, indicating that advanced signal processing techniques are required for bogie-based and car body-based wheel flat detection systems. The DIs showed clear differences between healthy and faulty wheelset and bogie frame acceleration signals, which is promising for eventual on-board wheel flat detection systems.

A Novel CBR-Decision Tree Based Intelligent Car Fault Diagnosis System (CFDS)

Fault diagnostic systems find numerous applications in almost all service domains. Now a days, interest lies on intelligent fault diagnosis. For this, a huge database of cases called Case Base (CB) comprising of fault descriptions and their solutions needs to be maintained. Case Based Reasoning (CBR) is a popular Artificial Intelligence technique that supports huge databases that find popular applications in fault diagnosis systems. CBR is a very useful process for solving problems, detecting and diagnosing faults, learning, reasoning and supporting decisions. As the size of CB increases, the accuracy of the system too increases leading to an increase in computational time complexity. So CBR techniques are coupled with machine learning approaches to reduce the same. This paper proposes a CBR methodology based Intelligent Car Fault Diagnosis System (CFDS) that integrates decision tree as a machine learning technique and jaccard similarity method to diagnose faults of cars accurately in a minimum time. A car fault diagnosis and detection system requires individual expertise gathered from personal experience and technical skills. Many times, not only the fault, but also the car part from where the fault originates or the cause of the fault needs to be known to handle or repair the problem. So, to help car mechanics as an assistant tool, CFDS is proposed; so that they can deal with various types of car faults very easily. Here, the proposed methodology integrates decision trees and jaccard similarity method to diagnose faults where the usage of decision trees is to store cases and jaccard similarity method is used to calculate the similarity percentages between user new query case and stored cases in the CB. User can post a new query about his car problem to the user interface of the CFDS. The CFDS uses the proposed methodology to find the solutions of that problem, and finally then at last these solutions are displayed to the user. To obtain better performance of the CFD system, this paper introduces a novel model of CBR cycle called CR4 model that is slightly modified version of traditional CBR cycle of R4 model, proposed by Aamodt and Plaza in 1994.

Capsule Networks for Object Detection in UAV Imagery

Recent advances in Convolutional Neural Networks (CNNs) have attracted great attention in remote sensing due to their high capability to model high-level semantic content of Remote Sensing (RS) images. However, CNNs do not explicitly retain the relative position of objects in an image and, thus, the effectiveness of the obtained features is limited in the framework of the complex object detection problems. To address this problem, in this paper we introduce Capsule Networks (CapsNets) for object detection in Unmanned Aerial Vehicle-acquired images. Unlike CNNs, CapsNets extract and exploit the information content about objects’ relative position across several layers, which enables parsing crowded scenes with overlapping objects. Experimental results obtained on two datasets for car and solar panel detection problems show that CapsNets provide similar object detection accuracies when compared to state-of-the-art deep models with significantly reduced computational time. This is due to the fact that CapsNets emphasize dynamic routine instead of the depth.

Abstract 2324: Preclinical safety, biodistribution, and tumor infiltration analysis of CAR T cell targets using in situ hybridization technology

Abstract Chimeric antigen receptor (CAR) T cell therapy has proven to be highly effective in treating hematologic malignancies, and major efforts are being made to achieve similar efficacy in solid tumors. These efforts face multiple challenges, including off-tumor target expression and checkpoint inhibition of CAR T cell activity in the tumor microenvironment. CAR T cells are much more potent compared to antibody therapeutics, therefore there is a need for more stringent CAR T target safety assessment to avoid adverse events resulting from “on-target/off-tumor” activity. Furthermore, it is critical to track and monitor CAR+ T cells within the context of intact tissue and tumor to understand the mechanisms underlying off-tumor toxicity and efficacy in tumor killing. In this study we employed the RNAscope in situ hybridization (ISH) technology to assess target expression specificity and to track CAR T cell distribution and activation in xenograft and host tissues using the RPMI-8226 xenograft mouse model, which expresses both BCMA and ROR1. The anti-ROR1 CAR T cells used in this study recognized both the mouse and human proteins whereas the anti-BCMA CAR T cells recognized only the human protein. RNA ISH revealed that BCMA was only expressed in the xenograft tumor and in no mouse organs, while ROR1 was found to be expressed in the xenograft tumor as well as at low levels in mouse lung and liver. Duplex RNA ISH assay with probes targeting the CAR 3’ UTR and either IFNG or GZMB mRNA allowed highly sensitive and specific detection of CAR T cells and their activation state in both tumor and normal tissues from vehicle, anti-ROR1 CAR T cell, or anti-BCMA CAR T cell treated mice. Activated anti-BCMA CAR T cells expressing GZMB and IFNG were found only in the xenograft tumor, where BCMA was expressed. In contrast, activated anti-ROR1 CAR T cells were found almost exclusively in mouse lung and liver with very few anti-ROR1 CAR T cells being found in the xenograft tumor, consistent with the previously observed pulmonary and hepatic toxicity of anti-ROR1 CAR T cells that was not predicted by IHC analysis of ROR1 protein presumably due to lack of antibody sensitivity. Lastly, we employed a multiplex ISH-IHC approach to confirm the presence of activated anti-BCMA CAR T cells in the xenograft tumor through simultaneous detection of the BCMA CAR 3’ UTR, IFNG, GZMB, and CD3, validating the anti-tumor activity of anti-BCMA CAR T cells. These data thus demonstrate how the RNAscope ISH assay can be utilized for CAR T efficacy and safety/toxicity assessment in preclinical models by detecting very low levels of target antigen expression in off-tumor tissues and monitoring CAR T cell pharmacodynamics and activation in tumor models. This technology has equal utility in understanding both CAR T and TCR-T cell activity in patient tumors. Citation Format: Helly Pimentel, Helen Jarnagin, Hailing Zong, Courtney Todorov, Kenneth Ganley, Fay Eng, Kevin Friedman, Molly Perkins, Shannon Grande, Courtney M. Anderson, Bingqing Zhang, Christopher Bunker, James B. Rottman, Xiao-Jun Ma. Preclinical safety, biodistribution, and tumor infiltration analysis of CAR T cell targets using in situ hybridization technology [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 2324.