Foggy Research Articles

Geryon

Vision-based drone-view object detection suffers from severe performance degradation under adverse conditions (e.g., foggy weather, poor illumination). To remedy this, leveraging complementary mmWave radar has become a trend. However, existing fusion approaches seldom apply to drones due to i) the aggravated sparsity and noise of point clouds from low-cost commodity radars, and ii) explosive sensing data and intensive computations leading to high latency. To address these issues, we design Geryon, an edge assisted object detection system on drones, which utilizes a suit of approaches to fully exploit the complementary advantages of camera and mmWave radar on three levels: (i) a novel multi-frame compositing approach utilizes camera to assist radar to address the aggravated sparsity and noise of radar point clouds; (ii) a saliency area extraction and encoding approach utilizes radar to assist camera to reduce the bandwidth consumption and offloading latency; (iii) a parallel transmission and inference approach with a lightweight box enhancement scheme further reduces the offloading latency while ensuring the edge-side accuracy-latency trade-off by the parallelism and better camera-radar fusion. We implement and evaluate Geryon with four datasets we collect under foggy/rainy/snowy weather and poor illumination conditions, demonstrating its great advantages over other state-of-the-art approaches in terms of both accuracy and latency.

IRDCLNet: Instance Segmentation of Ship Images Based on Interference Reduction and Dynamic Contour Learning in Foggy Scenes

Frequent bad weather at sea severely damages the quality of visual images captured by imaging equipment. Ship instance segmentation in adverse weather conditions remains a major challenge because of poor visibility at sea. Existing approaches for instance segmentation are primarily designed for clear days and rarely consider the aforementioned severe weather. Blurred ship objects can easily cause missed ship detection and decrease the instance segmentation performance on ship images, especially in the case of frequent fog at sea. To this end, we propose a ship instance segmentation framework (IRDCLNet) based on Interference Reduction and Dynamic Contour Learning in foggy scenes. The Interference Reduction Module is proposed to reduce the interference caused by fog and solves the problem of missed ship detection. Meanwhile, we present Dynamic Contour Learning to predict the overall contour of the blurred ships to assist in mask prediction. To handle the scarcity of ocean data in foggy weather, we build the Foggy ShipInsseg dataset, which contains 5,739 real and simulated foggy ship images with 10,900 fine instance mask annotations. Experiments on the Foggy ShipInsseg dataset show that our IRDCLNet outperforms the Mask R-CNN and CondInst baselines and achieves the state-of-the-art performance.

A Novel Detection Method Using YOLOv5 for Vehicle Target under Complex Situation

Aiming at the problem of low accuracy of vehicle image detection and recognition caused by low visibility in foggy weather, an improved YOLOv5 algorithm is proposed. This algorithm adjusts the brightness and contrast of the image by adding the improved adaptive histogram equalization method to the image preprocessing, highlights the detailed information of vehicle image signs, and changes the backbone network standard convolution mode to the depth separable convolution method for model lightweight processing. By constructing the corresponding vehicle target detection data set, this paper is superior to the object detection model commonly used on the public data set in terms of performance and effectiveness, and draws the following conclusion from the comparison results of ablation experiments, the improved algorithm improves the detection accuracy of a single image and reduces the processing time.

Software-based simulation to analyze the variation of digital modulation and atmospheric condition on the free space optic (FSO) link performance

Free Space Optic (FSO) is the solution for telecommunications technology that offers high data rates, wide bandwidth, and low power consumption. However, to maximize the performance of the FSO system, the modulation used should be considered in environmental conditions. This study aims to compare the performance of the FSO communication link based on digital modulation variations used in various weather conditions, including sunny, rainy, and foggy weather. This study uses two attenuation models, namely the Kim and Kruse models, with variations in transmission distance from 500 meters to 10 kilometers. Modulation variations used include QPSK, 8-PSK, 16-PSK, and 16-QAM at 10 Gbps bitrate. The simulation is accomplished using OptiSystem 17.0 software. The study results show that sunny weather (very clear) has the best visibility compared to rain and fog conditions with an attenuation value of 0.46 dB/km on the Kim and Kruse models. QPSK modulation has the best performance with a BER value of less than 1x10-12 up to a transmission distance of 8 km in sunny weather, 3 km in rainy weather (medium rain), and 800 m in foggy (moderate fog) weather. The 8-PSK modulation has a BER value of less than 1x10-12 with a range of 2000 m in sunny weather and 1500 m in rainy weather but does not meet the standards in foggy weather conditions. 16-PSK and 16-QAM modulation have above baseline BER values ​​during rainy and foggy conditions, but 16-QAM modulation still has a BER value of less than 1x10-3 during foggy conditions at a distance of 500 m.

Seasonal fog enhances crop water productivity in a tropical rubber plantation

• The effect of fog on rubber trees ( Hevea brasiliensis ) was quantified for 3 years. • Visibility, sap flux density and crop water productivity (WP c ) were studied. • Under the fog days, rubber trees were physiologically less active. • Carbon assimilation and evapotranspiration were suppressed but WP c enhanced by fog. • Dry seasonal fog can reduce water demand and alleviate drought stress. The rapid conversion of tropical rainforests into monoculture plantations of rubber ( Hevea brasiliensis ) in Southeast Asia (SEA) necessitates understanding of rubber tree physiology under local climatic conditions. Frequent fog immersion in the montane regions of SEA may affect the water and carbon budgets of the rubber trees and the plantation ecosystems. We studied the effect of fog on various plant physiological parameters in a mature rubber plantation in southwest China over 3 years. During the study period, an average of 141 fog events occurred every year, and the majority occurred during the dry season, when the temperature was relatively low. In addition to the low temperature, fog events were also associated with low vapor pressure deficit, atmospheric water potential, relative humidity and frequent wet-canopy conditions. We divided the dry season into cool dry (November-February) and hot dry (March-April) seasons and classified days into foggy (FG) and non-foggy (non-FG) days. During the FG days of the cool dry season, the physiological activities of the rubber trees were suppressed where carbon assimilation and evapotranspiration showed reductions of 4% and 15%, respectively, compared to the cool dry non-FG days. Importantly, the unequal declines in carbon assimilation and evapotranspiration led to enhanced crop water productivity (WP c ) on cool dry FG days but insignificant WP c values were found between FG and non-FG days of the hot dry season. Our results suggest that, by regulating plant physiology, fog events during the cool dry season significantly reduce water demand and alleviate water stress for the trees through improved WP c .

An automated extraction of small- and middle-sized rice fields under complex terrain based on SAR time series: A case study of Chongqing

Spatial assessment of rice cultivation area is a crucial activity, which is the foundation for the government to effectively improve the comprehensive rice production, promote the continuous increase of farmers' income and accelerate the construction of a modern rice industry system. In recent decades, large areas of arable land in Chongqing were converted to built-up land or restored to forests and grasses, resulting in a rapid decline in the rice planting area. Chongqing has an annual average of 104 cloudy and foggy days, the terrain is complex, and rice fields are generally fragmented with small- or middle-sized. How to timely obtain updated and accurate rice planting maps in Chongqing remains challenging. In this study, we processed 411 time series Sentinel-1A SAR scenes of vertical-horizontal polarization in 2020 over Chongqing, and analyzed the characteristics of rice phenological parameters under different terrain conditions, including rice transplanting date, mature grain date, length of growing season, rice agronomy flooding decline speed during sowing-transplanting period, and green-up speed during transplanting-mature period. On these bases, a decision tree algorithm integrating topographical features and rice phenological characteristics was proposed for collectively mapping patches of rice field. The identified rice results had user, producer, and overall accuracies of 0.96, 0.85 and 0.88, respectively. Meanwhile, SAR-derived rice area was compared against official rice statistical area at the county/district level with the correlation coefficient of 0.96. According to the rice map in 2020, there was a total area of 6125.70 km2 paddy rice in Chongqing, which was 6.27% lower than the data from the Chongqing statistical yearbook. Our study demonstrates the robustness and effectiveness of the proposed rice mapping method that comprehensively deliberate the topographical and rice phenological features in the complex landscapes with diverse crop types, small-medium sized and fragmented rice fields, and frequent cloudy and foggy weather.

Enhanced Performance of the 4 × 20 Gbit/s-40 GHz OFDM-Based RoFSO Transmission Link Incorporating WDM-MDM of Hermite Gaussian and Laguerre Gaussian Modes

A 4 × 20 Gbit/s-40 GHz orthogonal frequency-division multiplexing-based radio-on-free-space optics information link, integrating hybrid wavelength- and mode-division multiplexed transmission techniques, is proposed in this paper. Furthermore, an ameliorated detection scheme deploying a square root module after each photo diode at the receiver terminal is proposed to mitigate the adverse performance effects because of nonlinear photodiode characteristics. The presented results show a successful 80 Gbit/s-160 GHz transmission at 3000 and 2700 m under heavy rainy and foggy weather, respectively. The link demonstrates superior performance in terms of maximum range and transmission rate compared to previous works.

In-vehicle warnings for work zone and related rear-end collisions: A driving simulator experiment

Work zone area on roads is a critical component of road networks which concerns the safety of workers and passing by drivers. However, the passive speed reduction and lane changes caused by lane closure have led to frequent rear-end collisions in work zone areas. To help drivers better anticipate work zone situation and reduce collision risks, this paper proposes two types of in-vehicle warnings for work zone areas: Leading Vehicle Brake Warning (LVBW), and Lane-Closed Warning & Leading Vehicle Brake Warning (LCW & LVBW). The LVBW delivers a danger warning message to drivers upon the brake of the leading vehicle, while the LCW & LVBW provides an additional work-zone position message to remind drivers to decelerate in advance. A driving simulator experiment was conducted with 44 participants (24 males and 20 females) to test drivers’ performance in work zone area under different conditions, comprising two warning types (LVBW vs. LCW & LVBW), four warning times (3 s, 5 s, 7 s and 9 s) and two visibility conditions (clear and foggy weather). The results showed significant safety benefits of the lane-closed warning message under the LCW & LVBW condition. In contrast, the warning of leading vehicle’s brake in both LVBW and LCW & LVBW conditions had limited efficacy, which indicates that earlier warning about lane-closure is important to assist drivers in anticipating the complex situations in work zones. Drivers’ speed control and collision avoidance performances were impaired in fog, but the impairment was compensated by the warning messages. Compared with male drivers, female drivers tend to be more cautious when approaching the work zone areas. Overall, this study plays a pioneering role in developing effective safety countermeasures for work zone areas and providing strong support for implementing in-vehicle warning technologies.

Insulator Umbrella Disc Shedding Detection in Foggy Weather.

The detection of insulator umbrella disc shedding is very important to the stable operation of a transmission line. In order to accomplish the accurate detection of the insulator umbrella disc shedding in foggy weather, a two-stage detection model combined with a defogging algorithm is proposed. In the dehazing stage of insulator images, solving the problem of real hazy image data is difficult; the foggy images are dehazed by the method of synthetic foggy images training and real foggy images fine-tuning. In the detection stage of umbrella disc shedding, a small object detection algorithm named FA-SSD is proposed to solve the problem of the umbrella disc shedding occupying only a small proportion of an aerial image. On the one hand, the shallow feature information and deep feature information are fused to improve the feature extraction ability of small targets; on the other hand, the attention mechanism is introduced to strengthen the feature extraction network’s attention to the details of small targets and improve the model’s ability to detect the umbrella disc shedding. The experimental results show that our model can accurately detect the insulator umbrella disc shedding defect in the foggy image; the accuracy of the defect detection is 0.925, and the recall is 0.841. Compared with the original model, it improved by 5.9% and 8.6%, respectively.

Injury severity assessment of rear-end crashes via approaches based on generalized estimating equations

Rear-end crashes constitute the predominant type of crashes on highways and may lead to severe injuries and high property damage. Available statistical models primarily focus on injury severity and analyze potential factors that affect it. However, rear-end crashes may also be potentially correlated to vehicle, roadway, environmental, temporal, spatial, traffic, and crash characteristics. Additionally, unobserved heterogeneity regarding the effects may be present, which may be different in different crashes. In this context, multiple generalized estimating equation (GEE)-based models, developed using different working matrices and distributions, are proposed in this study to examine factors that affect injury severity. The proposed models account for both crash-related correlations and unobserved heterogeneity, thereby outperforming traditional models in terms of prediction accuracy. Among the explanatory variables considered in this study, the passenger car, minibus, curvature ratio, rainy weather, foggy weather, early morning, Thursday, autumn, winter, and average annual daily traffic volume were identified as contributing factors. However, significant differences were observed between the elasticity effects measured by different models, especially in terms of minibus and foggy weather. Thus, this study verifies that GEE-based models account for a greater amount of unobserved heterogeneity, yield better performance in terms of precision, and exhibit more consistent explanatory power compared to traditional models.

SJDL-Vehicle: Semi-supervised Joint Defogging Learning for Foggy Vehicle Re-identification

Vehicle re-identification (ReID) has attracted considerable attention in computer vision. Although several methods have been proposed to achieve state-of-the-art performance on this topic, re-identifying vehicle in foggy scenes remains a great challenge due to the degradation of visibility. To our knowledge, this problem is still not well-addressed so far. In this paper, to address this problem, we propose a novel training framework called Semi-supervised Joint Defogging Learning (SJDL) framework. First, the fog removal branch and the re-identification branch are integrated to perform simultaneous training. With the collaborative training scheme, defogged features generated by the defogging branch from input images can be shared to learn better representation for the re-identification branch. However, since the fog-free image of real-world data is intractable, this architecture can only be trained on the synthetic data, which may cause the domain gap problem between real-world and synthetic scenarios. To solve this problem, we design a semi-supervised defogging training scheme that can train two kinds of data alternatively in each iteration. Due to the lack of a dataset specialized for vehicle ReID in the foggy weather, we construct a dataset called FVRID which consists of real-world and synthetic foggy images to train and evaluate the performance. Experimental results show that the proposed method is effective and outperforms other existing vehicle ReID methods in the foggy weather. The code and dataset are available in https://github.com/Cihsaing/SJDL-Foggy-Vehicle-Re-Identification--AAAI2022.

Machine and Deep Learning Techniques for Daytime Fog Detection in Real Time with In-Vehicle Vision Systems Using the SHRP 2 Naturalistic Driving Study Data

The main focus of this study is to develop a system that can accurately detect the presence of fog in real time at a trajectory level. This study leveraged video data from the SHRP 2 Naturalistic Driving Study (NDS). Extensive data reduction steps were taken to classify various levels of foggy weather conditions from the video data to form two unique image data sets. Afterward, features based on the gray level co-occurrence matrix (GLCM) were extracted from the images and used as classification parameters for training support vector machine (SVM) and K-nearest neighbor (K-NN) algorithms. In addition, a convolutional neural network (CNN) was also examined to improve the detection performance. Although the analysis was done initially on a data set consisting of two weather conditions, clear and fog, it has been extended to include different levels of fog, that is, near fog and distant fog. While the accuracy of the first analysis with two categories was approximately 92% and 91% for SVM and K-NN classifiers, respectively, the CNN produced much greater accuracy of 99%. As expected, the accuracy of the second analysis, with more refined weather categories, was relatively lower than the first analysis where CNN, SVM, and K-NN models produced an accuracy of about 98%, 89%, and 88%, respectively. With the rapid advances in connectivity and affordable cameras, the proposed detection models could be integrated into the smartphones of regular road users, creating an effective way to collect real-time road weather information that could be used to improve weather-based variable speed limit (VSL) systems.

ObjectDetect: A Real-Time Object Detection Framework for Advanced Driver Assistant Systems Using YOLOv5

In the recent times, there has been a lot of speculation related to advanced driver-assistance system (ADAS) which provides best driving experience for the drivers. ADAS technology helps to detect the unhealthy driving conditions which lead to road accidents today. Road accidents are basically caused due to distracted driving, over speeding, drink and drive, foggy weather, no proper headlights, or due to some object which suddenly trespasses the vehicle. Today the major advancements in ADAS include parking assistance, road traffic detection, object detection on highways, and lane detection. But the major risk limitation in ADAS system is the speed and time at which the object is detected and tracked. Several algorithms such as R-CNN, Fast R-CNN, and YOLO were used for effective object detection and tracking earlier, but sometimes, the system do fail to detect due the speed factor. Hence, the proposed work presents a novel approach called “A Real-Time Object Detection Framework for Advanced Driver Assistant Systems” by implementing the state-of-the-art object detection algorithm YOLOv5 which improves the speed in detection of object over real-time. This paper provides a comparison between other state-of-the-art object detectors such as YOLOv3 and YOLOv4. Comparison is done based on mean average precision (mAP) and frames per second (FPS) on three benchmark datasets collected as a part of research findings. YOLOv5 proves to be faster and 95% accurate than the other object detection algorithms in the comparison. This framework is used to build a mobile application called “ObjectDetect” which helps users make better decisions on the road. “ObjectDetect” consists of a simple user interface that displays alerts and warnings.

Ankara Esenboga International Airport (Turkey) fog analysis and synoptical investigation of the fog event dated 17–19 December 2019

AbstractThe aviation industry has developed greatly in recent years. Millions of people choose air transport as a means of transportation every year, and flight safety is critically important. However, meteorological conditions can cause disruptions in flight activities and lead to accidents. One of the meteorological conditions that adversely affect flight activities is fog. Prediction of fog incidents increases flight safety. In this study, fog analysis of Ankara Esenboga International Airport was carried out between 2011 and 2020. The data used in the fog analysis are taken from the observations of Aviation Routine Weather Report (METAR) and Aviation Selected Special Weather Report (SPECI). The distribution of fog events in a 10‐year period was determined according to years, months, days, hours, meteorological parameters (wind, pressure, cloud cover/height, temperature, and prevailing view), and types. Over a 10‐year period, 802.02 hr of fog occurred. The average number of foggy days was determined as 17.9. According to the fog formation types, the fogs occurring at Esenboga Airport were classified as 62% radiation fog, 14% lowering cloud base, 9% precipitation fog, 2% advection fog, and 12% unknown fog. CAT categories in the foggy hours of Esenboga Airport were determined. In cases where there were foggy observations for approach and landing activities during the 10‐year period, these were found to be 40.34% according to CAT I, 72.8% according to CAT II, 93.8% according to CAT IIIA, and 100% according to CAT IIIB. Synoptic analysis of a particular fog incident, which started at 2220 UTC on December 16 and continued for 55.83 hr, was also made. When the synoptic maps and Skew‐T Log P diagrams were examined, the deepening of the high pressure center in and around Ankara and the low thermal advection caused a very stable atmosphere, resulting in a prolonged and intense fog event.

A strip line technique based 1 Gb/s, 70-dB linear dynamic range transimpedance amplifier towards LiDAR unmanned vehicle application

1This work reports a Microstrip Line (MSL) based Dual-Gate MOSFET (DGMOSFET) Transimpedance amplifier (TIA) for LIDAR unmanned vehicle application under different weather conditions. TIA (Design-I) is proposed under normal weather, while TIA (Design-II) for foggy weather conditions. TIA (Design-I) employs a variable gain common gate topology with post-amplification, resulting in high gain, wide bandwidth, and high dynamic range (DR). TIA (Design-II) incorporates a series MSL section at the input of TIA (Design-I), which further enhances the bandwidth performance. TIA (Design-I) realizes a fractional bandwidth of 104.3% with a transimpedance gain of 100.4 dBΩ and low input-referred noise (IRN) density of 4.29 pA/√Hz. TIA (Design-II) achieves a fractional bandwidth of 178.4% with transimpedance gain, IRN, and DR of 100.42 dBΩ, 3.81 pA/√Hz, and 70 dB, respectively. TIA (Design-II) demonstrates a 1 Gb/s data rate with a bit-error-rate < 10−10. The TIA (Design-I) and TIA (Design-II) consume the power of 33 mW and 39 mW under the supply voltage of 2 V.

Vehicle detection in foggy weather based on an enhanced YOLO method

Vehicle detection is the key to driverless technology. For safety, driverless technology requires extremely high accuracy and real-time for vehicle detection in different situations. In this paper, we study an enhanced YOLO -based algorithm for vehicle detection in foggy weather conditions. We add a dehazing module in the YOLO model for more information restoration, which is built by the multi-scale retinex with color restoration (MSRCR). And the enhanced model is trained with the augmentation data processed MSRCR for more stable performance. We evaluate our method in the public dataset, the results show the enhanced YOLO model has better performance than conventional YOLO in vehicle detection in foggy weather.

Performing the high bitrate visible light communications in the foggy weather to anticipate the interference on vehicle communications

We propose transmission media with visible light to communicate between vehicles. We evaluate the research under four scenarios using the modulation of On-Off Keying Non-Return to Zero (OOK-NRZ) and bitrate up to 1 Gbps. These scenarios are (i) ideal conditions, (ii) interference from other vehicle lights, (iii) foggy conditions, and (iv) interference from vehicles and fog conditions. Based on the extensive simulation, the results obtained are that interference and fog conditions can affect and reduce the value of the signal to interference and noise ratio (SNR) and increase the value of the error rate (BER). The results obtained are that interference and fog conditions can affect and reduce the value of the signal to interference and noise ratio (SNR) and increase the value of the error rate (BER). The SNR value in the first scenario is 23.6 dB and the second scenario is 11.1 dB, where this value is still sufficient. The SNR in scenario three is 16.1 dB, and the lowest in the fourth scenario is -7.78326 dB, indicating that the noise is extensive compared to signal power. In addition, we also obtain an optimal distance of communication between vehicles for each scenario sequentially 14.5 m, 13 m, 11.5 m, and 9 m.

Performance analysis of quantum key distribution using polarized coherent-states in free-space channel

In free space channel, continuous-variable quantum key distribution (CV-QKD) using polarized coherent-states can not only make the signal state more stable and less susceptible to interference based on the polarization non-sensitive of the free-space channel, but also reduce the noise introduced by phase interference. However, arbitrary continuous modulation can not be carried out in the past polarization coding, resulting in that the signal state can not obtain arbitrary continuous value in Poincare space, and the security analysis of CV-QKD using polarized coherent-states in free space is not complete. Here we propose a new modulation method to extend the modulation range of signal states with an optical-fiber-based polarization controller. In particular, in terms of the main influence factors in the free-space channel, we utilize the beam extinction and elliptical model when considering the transmittance and adopt the formulation of secret key rate. In addition, the performance of the proposed scheme under foggy weather is also taken into consideration to reveal the influence of severe weather. Numerical simulation shows that the proposed scheme is seriously affected by attenuation under foggy weather. The protocol fails when visibility is less than 1 km. At the same time, the wavelength can affect the performance of the proposed scheme. Specifically, under foggy weather, the longer the wavelength, the smaller the attenuation coefficient, and the better the transmission performance. Our proposed scheme can expand the modulation range of signal state, and supplement the security research of the scheme in the free-space channel, thus can provide theoretical support for subsequent experiments.

Investigation on variable speed limit control strategy of expressway under adverse weather conditions

In order to effectively improve the capacity of expressways under adverse weather conditions (rainy, foggy and snowy) and reduce the potential safety hazards of expressways, a variable speed limit control strategy is proposed. First, the impact of road alignment on traffic capacity is considered, and road traffic capacity reduction coefficients under different bends and slopes are added so that the cell transmission model and a dynamic traffic flow model can be established; Secondly, the maximum speed limit value under adverse weather conditions is determined according to the influence of rainfall intensity under rainy weather conditions, visibility under foggy weather conditions, and road adhesion coefficient under snowy weather conditions on traffic capacity; Finally, an adaptive genetic algorithm is employed to solve the function that satisfies the safety and efficiency of traffic operation. The results show that the variable speed limit control strategy under adverse weather conditions effectively improves the road capacity, reduces the driving risk of vehicles, and alleviates traffic congestion.

Importance of 3D convolution and physics on a deep learning coastal fog model

The forecasting of hazardous atmospheric phenomena is often challenging. Artificial intelligence (AI) models have been applied to atmospheric science problems. Model complexity provides a motivation to quantify the importance of model architecture components. We studied the relative importance of the components of the FogNet model that was designed for big atmospheric data: 1) 3D versus 2D convolution, 2) physics-based grouping and ordering of meteorological input features, 3) different auxiliary CNN-based feature learning modules and 4) parallel versus sequential spatial-variable-wise feature learning. We investigate the relative importance of these CNN architectural features by predicting coastal fog,a complex spatiotemporal dynamical process. We use four explainable AI techniques to better understand input feature contributions. The results of the experiments demonstrate that 3D-CNN based models better capture the complexity of the fog prediction process than the 2D-CNNs. We also show that physics-based feature grouping, and the order in which they are fed into the CNNs, significantly impacts performance.