Complex Background Scenes Research Articles

Pixel-Level Crack Identification for Bridge Concrete Structures Using Unmanned Aerial Vehicle Photography and Deep Learning

Traditional manual inspection technology has the problems of high risk, low efficiency, and being time-consuming in bridge safety management. The unmanned aerial vehicle (UAV)-based detection technology is widely used in bridge structure safety monitoring. However, the existing deep learning-based concrete crack identification method has great limitations in dealing with complex background and tiny cracks in bridge structures. To address these problems, this study designs a crack pixel-level high-performance segmentation model for bridge concrete cracks that is suitable for UAV detection scenarios using machine vision (MV) and deep learning (DL) algorithms. First, considering the high requirements for the computing performance of the MV-based model for UAV-based detection, the ResNet-18-based lightweight convolutional neural network is used to represent the traditional large-scale backbone network of the pyramid scene parsing network (PSPNet) to develop a high-performance crack automatic identification model. Then, considering that bridge concrete cracks have the characteristics of subtle shapes and complex backgrounds, the spatial position self-attention module is inserted into the PSPNet to improve its detection accuracy. A concrete bridge is used for the case study, and a dataset of cracks in bridge concrete structures collected by UAVs is constructed and used for model training. The experimental results show that the loss function of the developed method in the training process results in a smooth decline, and the developed algorithm achieves the evaluation indicators of 0.9008 precision, 0.8750 recall, 0.8820 accuracy, and 0.9012 IOU on the bridge concrete crack dataset, which are significantly higher than other state-of-the-art baseline methods. In addition, four common UAV bridge detection scenarios, including low light, complex crack forms, high background roughness, and complex background scenes, are used to further test the crack detection ability of the developed crack identification model. The experimental results show that the proposed crack identification method can effectively overcome interference and real-size pixel-level segmentation of crack morphology. In addition, it also achieved a detection efficiency of 35.04 FPS, which shows that the real-time detection ability of the method has good applicability in the UAV detection scene.

Recent Advances in Traffic Sign Recognition: Approaches and Datasets

Autonomous vehicles have become a topic of interest in recent times due to the rapid advancement of automobile and computer vision technology. The ability of autonomous vehicles to drive safely and efficiently relies heavily on their ability to accurately recognize traffic signs. This makes traffic sign recognition a critical component of autonomous driving systems. To address this challenge, researchers have been exploring various approaches to traffic sign recognition, including machine learning and deep learning. Despite these efforts, the variability of traffic signs across different geographical regions, complex background scenes, and changes in illumination still poses significant challenges to the development of reliable traffic sign recognition systems. This paper provides a comprehensive overview of the latest advancements in the field of traffic sign recognition, covering various key areas, including preprocessing techniques, feature extraction methods, classification techniques, datasets, and performance evaluation. The paper also delves into the commonly used traffic sign recognition datasets and their associated challenges. Additionally, this paper sheds light on the limitations and future research prospects of traffic sign recognition.

CAE-Net: Cross-Modal Attention Enhancement Network for RGB-T Salient Object Detection

RGB salient object detection (SOD) performs poorly in low-contrast and complex background scenes. Fortunately, the thermal infrared image can capture the heat distribution of scenes as complementary information to the RGB image, so the RGB-T SOD has recently attracted more and more attention. Many researchers have committed to accelerating the development of RGB-T SOD, but some problems still remain to be solved. For example, the defective sample and interfering information contained in the RGB or thermal image hinder the model from learning proper saliency features, meanwhile the low-level features with noisy information result in incomplete salient objects or false positive detection. To solve these problems, we design a cross-modal attention enhancement network (CAE-Net). First, we concretely design a cross-modal fusion (CMF) module to fuse cross-modal features, where the cross-attention unit (CAU) is employed to enhance the two modal features, and channel attention is used to dynamically weigh and fuse the two modal features. Then, we design the joint-modality decoder (JMD) to fuse cross-level features, where the low-level features are purified by higher level features, and multi-scale features are sufficiently integrated. Besides, we add two single-modality decoder (SMD) branches to preserve more modality-specific information. Finally, we employ a multi-stream fusion (MSF) module to fuse three decoders’ features. Comprehensive experiments are conducted on three RGB-T datasets, and the results show that our CAE-Net is comparable to the other methods.

Improved SiamFC Target Tracking Algorithm Based on Anti-Interference Module

The SiamFC target tracking algorithm has attracted extensive attention because of its good balance between speed and performance, but the tracking effect of the SiamFC algorithm is not satisfactory in complex background scenes. When SiamFC algorithm uses deep semantic features for tracking, it has good recognition ability for different types of objects, but it has insufficient discrimination for the same types of objects. Therefore, we propose an effective anti-interference module to improve the discrimination ability of the algorithm. The anti-interference module uses another feature extraction network to extract the features of the candidate target images generated by the SiamFC main network. In addition, we set up the feature vector set to save the feature vectors of the tracking target and the template image. Finally, the tracking target is selected by calculating the minimum cosine distance between the feature vector of the candidate target and the vector in the feature vector set. A large number of experiments show that our anti-interference module can effectively improve the performance of SiamFC algorithm, and the performance of this algorithm can be comparable to the popular algorithms.

A Self-Attention Augmented Graph Convolutional Clustering Networks for Skeleton-Based Video Anomaly Behavior Detection

In this paper, we propose a new method for detecting abnormal human behavior based on skeleton features using self-attention augment graph convolution. The skeleton data have been proved to be robust to the complex background, illumination changes, and dynamic camera scenes and are naturally constructed as a graph in non-Euclidean space. Particularly, the establishment of spatial temporal graph convolutional networks (ST-GCN) can effectively learn the spatio-temporal relationships of Non-Euclidean Structure Data. However, it only operates on local neighborhood nodes and thereby lacks global information. We propose a novel spatial temporal self-attention augmented graph convolutional networks (SAA-Graph) by combining improved spatial graph convolution operator with a modified transformer self-attention operator to capture both local and global information of the joints. The spatial self-attention augmented module is used to understand the intra-frame relationships between human body parts. As far as we know, we are the first group to utilize self-attention for video anomaly detection tasks by enhancing spatial temporal graph convolution. Moreover, to validate the proposed model, we performed extensive experiments on two large-scale publicly standard datasets (i.e., ShanghaiTech Campus and CUHK Avenue datasets) which reveal the state-of-art performance for our proposed approach when compared to existing skeleton-based methods and graph convolution methods.

Experimental Evaluation of Computer Vision and Machine Learning-Based UAV Detection and Ranging

We consider the problem of vision-based detection and ranging of a target UAV using the video feed from a monocular camera onboard a pursuer UAV. Our previously published work in this area employed a cascade classifier algorithm to locate the target UAV, which was found to perform poorly in complex background scenes. We thus study the replacement of the cascade classifier algorithm with newer machine learning-based object detection algorithms. Five candidate algorithms are implemented and quantitatively tested in terms of their efficiency (measured as frames per second processing rate), accuracy (measured as the root mean squared error between ground truth and detected location), and consistency (measured as mean average precision) in a variety of flight patterns, backgrounds, and test conditions. Assigning relative weights of 20%, 40% and 40% to these three criteria, we find that when flying over a white background, the top three performers are YOLO v2 (76.73 out of 100), Faster RCNN v2 (63.65 out of 100), and Tiny YOLO (59.50 out of 100), while over a realistic background, the top three performers are Faster RCNN v2 (54.35 out of 100, SSD MobileNet v1 (51.68 out of 100) and SSD Inception v2 (50.72 out of 100), leading us to recommend Faster RCNN v2 as the recommended solution. We then provide a roadmap for further work in integrating the object detector into our vision-based UAV tracking system.

A Novel Approach for Detection of Moving Objects in Complex Scenes Using Fuzzy Colour Difference Histogram

Background subtraction schematic is widely used for motion detection. For effective automation of this process, a robust algorithm with high accuracy is needed. One of the major challenges of such algorithms is the identification of objects from an environment with composite elements that may be a dynamic background, frames with a camouflaged background and foreground pixels, and consecutive frames with varying illumination. The existing system uses a multi-color space histogram superposition principle having the biggest challenge of choosing appropriate color components in suitable proportion. Overcoming this challenge, a novel approach, MODITBS, processed in a differential domain, is proposed. A fuzzified color difference histogram-based background modeling is done to significantly deal with complex background scenes followed by principal component analysis-based feature extraction. The foreground objects detected are enhanced using a Kalman filter. The results show that MODITBS attains an accuracy of 95.16% in comparison to the existing system having an accuracy of 91.25%.

Level set method with Retinex‐corrected saliency embedded for image segmentation

It can be a very challenging task when using level set method segmenting natural images with high intensity inhomogeneity and complex background scenes. A new synthesis level set method for robust image segmentation based on the combination of Retinex-corrected saliency region information and edge information is proposed in this work. First, the Retinex theory is introduced to correct the saliency information extraction. Second, the Retinex-corrected saliency information is embedded into the level set method due to its advantageous quality which makes a foreground object stand out relative to the backgrounds. Combined with the edge information, the boundary of segmentation will be more precise and smooth. Experiments indicate that the proposed segmentation algorithm is efficient, fast, reliable, and robust.

Image Feature Extraction and Object Recognition Based on Vision Neural Mechanism

As an important branch of artificial intelligence, computer vision plays a huge role in the rapid development of artificial intelligence. From a biological point of view, in the acquisition and processing of information, vision is much more important than hearing, touch, etc., because 70% of the human cerebral cortex is processing visual information. Therefore, advances in computer vision technology are critical to the development of artificial intelligence that is designed to allow machines to think and handle things like humans. The acquisition and processing of visual information has always been the focus of computer vision research, and it is also difficult. The main problem of traditional computer vision technology in the processing of visual information is that the extracted image features are less discriminative, the generalization ability of image features in complex background scenes is insufficient, and the recognition ability on object recognition is poor. In response to these problems, based on the visual neural mechanism, this paper establishes an appropriate computer model for the neuronal cells in the human primary visual cortex, models the recognition response mechanism of the visual ventral system, and performs image feature extraction on the training samples. And object recognition. The results show that compared with the traditional methods, the proposed method effectively improves the discrimination of image features, and the image features extracted under complex background scenes have good generalization ability. On this basis, the training samples can be effectively recognized. The results show that the model based on the visual neural mechanism, the recognition of the edge, orientation and contour of the training sample show the advantages of the biological vision mechanism in object recognition.

RETRACTED ARTICLE: The visual object tracking algorithm research based on adaptive combination kernel

In order to enhance the robustness to complicated changes of multiple objects and complex background scene, the visual object tracking algorithm based on Adaptive Combination Kernel has been proposed in the paper. The object tracking procedure has been decomposed into two subtasks: Translation Filter and Scale Filter to estimate the object’s details. Firstly, the Translation Kernel Tracker has used the adaptive combination of Linear Kernel Filter and Gaussian Kernel Filter. The objective function has been developed to obtain the weight coefficients for Linear Kernel filter and the Gaussian Kernel filter, which incorporates not only empirical risk but also maximum value of response output for each kernel. The Adaptive Combination Kernel has the advantages of both local kernel and global kernel. Secondly, the tracking position has been calculated according to the response output of adaptive combination kernel correlation filter. Thirdly, according to the maximum response value, the scene-adaptive learning rate has been designed in the translation filter. The translation filter can be updated with the adaptive learning rate. Finally, one-dimensional scale filter has been used to estimate the object scale. The extensive experimental results have shown that the proposed algorithm is optimal on OTB-50 dataset in success rate and distance precision parameters, which is 6.8 percentage points and 4.1% points than those of KCF and is 2.0 percentage points and 3.2% points than those of BSET. The proposed algorithm has better robustness to the deformation and occlusion than others.

Visual salience and biological motion interact to determine camouflaged target detectability

Target visual salience and biological motion independently influence the accuracy and latency of observer detection. However, it is currently unknown how these target parameters might interact in modulating the detectability of camouflaged human targets. In two experiments, observers performed a visual target detection task. In a pilot experiment, observers detected a static human target with parametrically varied visual salience, superimposed on a complex background scene. As expected, results demonstrated varied target detectability as a function of salience, with observers showing higher hit rates and faster response times as a function of increased salience. In the Main Experiment, observers detected simulated human targets walking across a complex scene at five different speeds and three different levels of visual salience (as validated in the pilot experiment). We found strong effects of both movement rate and visual salience, and the two parameters interacted. Specifically, increasing the rate of biological motion increased detectability for even the least salient camouflage patterns. In other words, biological motion can “break” even the least conspicuous camouflage pattern. In contrast, a very salient pattern was highly detectable under static and moving conditions. Results are considered in relation to theories of camouflage detectability, and trade-offs between camouflage development efforts versus advanced training in military maneuvering.

Sign language recognition in complex background scene based on adaptive skin colour modelling and support vector machine

With the advances of wearable cameras, the user can record the first-person view videos for gesture recognition or even sign language recognition to help the deaf or hard of hearing people communicate with others. In this paper, we propose a purely vision-based sign language recognition system which can be used in complex background scene. We design an adaptive skin colour modelling method for hand segmentation so that the hand contour can be derived more accurately even when different users use our system in various light conditions. Four kinds of feature descriptors are integrated to describe the contours and the salient points of hand gestures, and support vector machine (SVM) is applied to classify hand gestures. Our recognition method is evaluated by two datasets: 1) the CSL dataset collected by ourselves in which images were captured in three different environments including complex background; 2) the public ASL dataset, in which images of the same gesture were captured in different lighting conditions. The proposed recognition method achieves acceptable accuracy rates of 100.0% and 94.0% for the CSL and ASL datasets, respectively.

Sign language recognition in complex background scene based on adaptive skin colour modelling and support vector machine

With the advances of wearable cameras, the user can record the first-person view videos for gesture recognition or even sign language recognition to help the deaf or hard of hearing people communicate with others. In this paper, we propose a purely vision-based sign language recognition system which can be used in complex background scene. We design an adaptive skin colour modelling method for hand segmentation so that the hand contour can be derived more accurately even when different users use our system in various light conditions. Four kinds of feature descriptors are integrated to describe the contours and the salient points of hand gestures, and support vector machine (SVM) is applied to classify hand gestures. Our recognition method is evaluated by two datasets: 1) the CSL dataset collected by ourselves in which images were captured in three different environments including complex background; 2) the public ASL dataset, in which images of the same gesture were captured in different lighting conditions. The proposed recognition method achieves acceptable accuracy rates of 100.0% and 94.0% for the CSL and ASL datasets, respectively.

Study on detection of preceding vehicles based on convolution neural network

A convolution neural network (ConvNet) based vehicle detection system is developed in view of this issue that vehicle detection based on monocular vision is susceptible to be disturbed by complex background scene. Firstly, in order to detect shadows underneath vehicles for generating the candidate regions of shadow underneath vehicle, a road detection method using edge enhancement as well as an adaptive shadow segmentation approach are applied, which are aimed to better deal with the problems of grayscale variation on road and reduce the impact of the lighting variance. Then the ConvNet’s structure applied to the road traffic environment is determined and trained by the established image sample sets. The shadow regions detected wrongly as the shadows underneath vehicles are recognized by ConvNet and removed from the preliminary detection results so as to precisely verify the presence of vehicles in an image. The experimental results indicate that this algorithm described in this paper is effective and precise, which can distinguish well between vehicles and background interferences.

Robust background subtraction to global illumination changes via multiple features-based online robust principal components analysis with Markov random field

Background subtraction is an important task for various computer vision applications. The task becomes more critical when the background scene contains more variations, such as swaying trees and abruptly changing lighting conditions. Recently, robust principal component analysis (RPCA) has been shown to be a very efficient framework for moving-object detection. However, due to its batch optimization process, high-dimensional data need to be processed. As a result, computational complexity, lack of features, weak performance, real-time processing, and memory issues arise in traditional RPCA-based approaches. To handle these, a background subtraction algorithm robust against global illumination changes via online robust PCA (OR-PCA) using multiple features together with continuous constraints, such as Markov random field (MRF), is presented. OR-PCA with automatic parameter estimation using multiple features improves the background subtraction accuracy and computation time, making it attractive for real-time systems. Moreover, the application of MRF to the foreground mask exploits structural information to improve the segmentation results. In addition, global illumination changes in scenes are tackled by using sum of the difference of similarity measure among features, followed by a parameter update process using a low-rank, multiple features model. Evaluation using challenging datasets demonstrated that the proposed scheme is a top performer for a wide range of complex background scenes.

Optimization of Face Recognition Algorithms for Smartphone Environment

The face recognition algorithm has been investigated as a robust solution for biometric authentication. It is used in a wide range of applications such as surveillance, access control, airport immigration system and application of user recognition in a smartphone, etc. (The security of face recognition system can be trusted, because they were used unique biometric data) Face recognition system tends to give a higher degree of recognition performance only when image acquisition process was restricted environment. In this paper, we explore face detection and face recognition algorithms for smartphone environment. The basic requirements include appropriate data format and collection of feature data to achieve efficiency of algorithm. Our experiment presents a face detection and recognition algorithm for smartphone environment. The essential part for proposed system includes; (1) integer representation from floating point calculation on desktop computer and smartphone, (2) optimization of memory management scheme and (3) efficient face detection performance on complex background scene by Modified Census Transform.

A Novel Object Tracking Based on Foreground Hue Histogram

We propose a tracking algorithm for a single non-rigid object based on its foreground hue histogram. A tracked region can be described by the foreground hue histogram only calculating foreground object pixels, which can effectively restrain the disturbing of complex background environments. For measuring the object likelihood, we match the foreground hue histogram with that of the tracked object and refer the result of motion detection to encircle the tracked object region as much as possible. During the tracking, we update the hue histogram model for adapting the object appearance variation. The proposed algorithm is realized in the particle filter frame, and the experiments show that it is capable of robustly and accurately tracking a single non-rigid object for the situations of complex background scenes and strong appearance variations.

Target Detection Algorithm Based on the Movement of Codebook Model

Movement target detection is the researchful emphasis and aporia in the fields of computer vision, model recognition and video coding. In order to extract moving targets from the complex background scenes, this paper puts forward to calculate the color distortion degree by means of getting every pixel or a group of pixels of time series model, and converting pixels from RGB space to HSV space. Based on the background Codebook model of the target motion detection algorithm, the experiment shows that the method can achieve better target detection quality.

An Improved Mean-Shift Tracking Algorithm using PSO-Based Adaptive Feature Selection

Abstract Traditional mean-shift tracking algorithm use pre-defined tracking feature. Its trends to lead tracking failure in the complex background scenes and fast-changing background scenes. In this paper, an improved mean-shift tracking algorithm using Particle swarm optimization (PSO) based adaptive feature selection is presented to improve the tracking performance. We assume that the features with best discrimination between object and background are also the best for tracking the object. A two-class variance ratio is employed to measure the discrimination. PSO algorithm is used to optimize the different feature combination to adaptively generate the best tracking feature. Experimental results show that the proposed method can improve the performance of mean-shift tracker significantly in the complex and fast-changing background scenes.

Automatic detection of dining plates for image-based dietary evaluation.

An automatic detector that finds circular dining plates in chronically recorded images or videos is reported for the study of food intake and obesity. We first detect edges from input images. After a number of processing steps that convert edges into curves, arc filtering and grouping algorithms are applied. Then, convex hulls are identified and the ones that fit the description of ellipses corresponding to dining plates are determined. Our experiments using real-world images indicate that this detector is highly reliable and robust even when the input images contain complex background scenes and the dining plates are severely occluded.