Few-Shot Learning Network for Moving Object Detection Using Exemplar-Based Attention Map

Moving Object Detection (MOD) is a fundamental step for many computer vision applications. MOD becomes very challenging when a video sequence captured from a static or moving camera suffers from the challenges: camouflage, shadow, dynamic backgrounds, and lighting variations, to name a few. Deep learning methods have been successfully applied to address MOD with competitive performance. However, in order to handle the overfitting problem, deep learning methods require a large amount of labeled data which is a laborious task as exhaustive annotations are always not available. Moreover, some MOD deep learning methods show performance degradation in the presence of unseen video sequences because the testing and training splits of the same sequences are involved during the network learning process. In this work, we pose the problem of MOD as a node classification problem using Graph Convolutional Neural Networks (GCNNs). Our algorithm, dubbed as GraphMOD-Net, encompasses instance segmentation, background initialization, feature extraction, and graph construction. GraphMOD-Net is tested on unseen videos and outperforms state-of-the-art methods in unsupervised, semi-supervised, and supervised learning in several challenges of the Change Detection 2014 (CDNet2014) and UCSD background subtraction datasets.

Moving object detection with a mobile image sensor is an important task in robotics and computer vision, when considering the practical use of robotics in human environments. In this paper, we propose a robust method that detects moving objects in the environment using the omnidirectional depth information obtained by a mobile Stereo Omnidirectional System (SOS). In order to detect only the moving objects within the depth image that are obtained by a sensor in motion, we first estimate the ego‐motion of the sensor, and generate a predicted depth image for the current time from the depth obtained at the previous time by only considering the ego‐motion of the sensor. Then the predicted depth image is compared with the actual one obtained at the current time, and the inconsistent regions are detected as moving objects. When the sensor moves, occlusions will occur in the scene and they will cause false detections. However, these false detections can be suppressed by estimating the occlusion regions using the ego‐motion parameters of the sensor and the jump edges in the depth image. The effectiveness of the method is shown with experimental results for a real environment. © 2005 Wiley Periodicals, Inc. Electr Eng Jpn, 152(3): 29–38, 2005; Published online in Wiley InterScience (www.interscience. wiley.com). DOI 10.1002/eej.20155

Moving object detection is an important research hotspot in the field of image processing, which is the basis of object classification and object tracking. On the one hand, moving object detection should adapt to the changes of environment and itself, on the other hand, it should take into account the accuracy and real-time of detection. ViBe algorithm is a simple and fast algorithm in moving object detection methods. It has no specific requirements for video format, color space, and background relationship. It only needs a small number of sample values to cover the background samples, and has both accuracy and computational efficiency. How to choose parameters is the key problem to determine the performance of ViBe algorithm. This paper mainly studies the ViBe detection method for moving objects, and optimizes the main parameters of the ViBe algorithm through simulation. Firstly, the sample radius is selected according to the target resolution; Secondly, the updating factor of time parameters and the number of samples in the pixel model are optimized; Finally, the minimum matching number is selected to constrain the number of matching samples to classify new pixels. The experimental results show that the method has good performance in moving object detection and can retain the integrity of moving objects.

Moving object detection under varying illumination is a prominent area of research in the field of computer vision. The detected object results from moving object framework, suffers from factors like noise, change in object scene, shadow, changes in illumination etc. So there is a growing interest to develop a robust moving object detection framework under varying illumination scenario. A significant amount of methods have been proposed over recent years on moving object detection under varying illumination condition. In this paper some of the recent effective methods based on accuracy have been reviewed. In this paper, methods reviewed include illumination sensitive feature and illumination invariant feature approaches for moving object detection under illumination varying background conditions. Methods reviewed in this paper include qualitative and quantitative comparison to guide the reader a clear insight to the computer vision community, those who are especially working in moving object detection under different illumination conditions.

Moving object detection in a video plays an important role in many vision applications. Recently, moving object detection using appearance modeling based on a convolutional neural network has been actively developed. However, the CNN- based methods usually require the user’s supervision of the first frame so that it becomes highly dependent on the training dataset. In contrast, the method of finding a foreground, which models a background occupying a large proportion in an image, can detect a moving object efficiently in an unsupervised manner. However, existing methods based on background modeling in a pan-tilt-zoom (PTZ) camera suffer many false positives or loss of moving objects due to the estimation error of camera motion. To overcome the aforementioned limitations, we propose a moving object detection method for a PTZ camera through two background models. In an unsupervised way, our method builds the two background models that have different roles: 1) a coarse background model for detecting large changes, and 2) a fine background model for detecting small changes. In more detail, the coarse background model builds a block-based Gaussian model, and the fine model builds a sample consensus model. Both models are adaptively updated according to the estimated camera motion in the video recorded by a PTZ camera. Then, each foreground result from two background models is incorporated to fill the moving object region. Through experiments, the proposed method achieves better performance than the state-of-the-art methods and operates in real-time without parallel processing. In addition, we showed the effectiveness of the proposed model through improved results of moving object detection through combination with the latest supervised method.

Video object segmentation is a challenging task in computer vision because the appearances of target objects might change drastically along the time in the video. To solve this problem, space-time memory (STM) networks are exploited to make use of the information from all the intermediate frames between the first frame and the current frame in the video. However, fully using the information from all the memory frames may make STM not practical for long videos. To overcome this issue, a novel method is developed in this paper to select the reference frames adaptively. First, an adaptive selection criterion is introduced to choose the reference frames with similar appearance and precise mask estimation, which can efficiently capture the rich information of the target object and overcome the challenges of appearance changes, occlusion, and model drift. Secondly, bi-matching (bi-scale and bi-direction) is conducted to obtain more robust correlations for objects of various scales and prevents multiple similar objects in the current frame from being mismatched with the same target object in the reference frame. Thirdly, a novel edge refinement technique is designed by using an edge detection network to obtain smooth edges from the outputs of edge confidence maps, where the edge confidence is quantized into ten sub-intervals to generate smooth edges step by step. Experimental results on the challenging benchmark datasets DAVIS-2016, DAVIS-2017, YouTube-VOS, and a Long-Video dataset have demonstrated the effectiveness of our proposed approach to video object segmentation.

Effective vehicle detection technique for traffic surveillance systems

Recent research has shown the initial success of sparse coding (Sc) in solving many computer vision tasks. Motivated by the fact that kernel trick can capture the nonlinear similarity of features, which helps in finding a sparse representation of nonlinear features, we propose kernel sparse representation (KSR). Essentially, KSR is a sparse coding technique in a high dimensional feature space mapped by an implicit mapping function. We apply KSR to feature coding in image classification, face recognition, and kernel matrix approximation. More specifically, by incorporating KSR into spatial pyramid matching (SPM), we develop KSRSPM, which achieves a good performance for image classification. Moreover, KSR-based feature coding can be shown as a generalization of efficient match kernel and an extension of Sc-based SPM. We further show that our proposed KSR using a histogram intersection kernel (HIK) can be considered a soft assignment extension of HIK-based feature quantization in the feature coding process. Besides feature coding, comparing with sparse coding, KSR can learn more discriminative sparse codes and achieve higher accuracy for face recognition. Moreover, KSR can also be applied to kernel matrix approximation in large scale learning tasks, and it demonstrates its robustness to kernel matrix approximation, especially when a small fraction of the data is used. Extensive experimental results demonstrate promising results of KSR in image classification, face recognition, and kernel matrix approximation. All these applications prove the effectiveness of KSR in computer vision and machine learning tasks.

Object detection is the core task in computer vision, widely used in industry, medicine, agriculture, and other fields. With the development of deep learning technology, object detection algorithms have made remarkable progress in improving accuracy, improving efficiency, and reducing cost. However, the existing technology still faces challenges such as insufficient accuracy, high computational complexity, and deployment flexibility. In this paper, the basic principle of object detection is summarized in detail, including two-stage and single-stage object detection algorithms, and the advantages and disadvantages of several classical algorithms and their application scenarios are analyzed. At the same time, this paper summarizes the specific cases of the application of object detection technology in industrial production, medical diagnosis, and smart agriculture, and discusses the limitations and future development direction of the existing object detection technology. The review in this paper provides a valuable reference for further enhancing the application potential of target detection technology, and it is expected that future research can overcome the current challenges and achieve a wider application.

Semantic segmentation is a core task in computer vision, but the high computational demands of state-of-the-art models hinder deployment in real-time and resource-constrained environments. Multi-exit architectures offer a potential remedy by enabling adaptive inference through early exits at intermediate layers, yet their application to dense prediction tasks remains underexplored. In this work, we extend multi-exit architectures to semantic segmentation using a SegNet backbone with two additional exits and systematically evaluate training strategies, including joint optimization, final exit distillation, and exit-ensemble distillation. Our experiments on the CamVid dataset show that knowledge distillation consistently improves both accuracy and stability, with the exit-ensemble strategy achieving the highest accuracy and final exit distillation offering the most stable convergence. We conduct a systematic evaluation of confidence metrics and identify Maximum Prediction as the most reliable estimator of prediction quality in our setting. Based on this, we apply a prediction recycling strategy that reuses intermediate outputs that would otherwise be discarded. This approach reduces computational waste and improves efficiency, achieving up to a 21.6% speedup over a single-exit baseline with only a 1.87 percentage-point drop in accuracy. These findings demonstrate the viability of multi-exit semantic segmentation for real-time applications and provide a principled foundation for future research on adaptive inference in dense prediction tasks.

Semantic segmentation is a core task in computer vision that allows AI models to interact and understand their surrounding environment. Similarly to how humans subconsciously segment scenes, this ability is crucial for scene understanding. However, a challenge many semantic learning models face is the lack of data. Existing video datasets are limited to short, low-resolution videos that are not representative of real-world examples. Thus, one of our key contributions is a customized semantic segmentation version of the Walking Tours Dataset that features hour-long, high-resolution, real-world data from tours of different cities. Additionally, we evaluate the performance of open-vocabulary, semantic model OpenSeeD on our own custom dataset and discuss future implications.

Single Image Super-Resolution (SISR) aims to reconstruct high-resolution (HR) images from low-resolution (LR) inputs, serving as a core task in computer vision. Despite recent advances, existing methods often struggle to balance structural fidelity and computational efficiency. To address this, we propose PFRNet, a lightweight superpixel-aware model integrating superpixel segmentation, local attention aggregation, and global structure modeling. The framework comprises four key modules: GASS, SPDF, SPFA, and LAP, jointly enabling multi-scale and structure-consistent feature learning. Experiments on benchmark datasets (e.g., Set5, Urban100) show that PFRNet achieves superior performance with fewer parameters. Ablation studies further verify the effectiveness of each module.

Contour extraction is a core task in computer vision, serving as the foundation for object detection, segmentation, and scene understanding across various applications, including autonomous vehicles, medical imaging, and industrial automation. This paper provides an in-depth review of both classical and modern artificial intelligence (AI)-based contour extraction models. The classical methods, such as edge detectors and gradient-based operators, alongside advanced AI models, including convolutional neural networks and semantic segmentation architectures are reviewed in it. By examining each model’s strengths, limitations, and applicability to diverse real-world tasks, we aim to provide a comprehensive guide to contour extraction, highlight key challenges, and propose potential research directions in this evolving field.

Moving objects have been detected using various object detection techniques. Two categories for moving object detection techniques are frame differencing based and background subtraction based. These techniques are limited by camera scene complexity, light conditions, video type etc. Frame differencing based techniques process videos faster compared to background subtraction based techniques. Frame differencing based techniques detects only the boundary of the moving object and may fail for slow moving objects. These techniques for moving object detection can be improved by using sound data as most video recording cameras are equipped with a microphone. Sounds from human footsteps can be recorded with video and used with frame differencing techniques to improve moving object detection results. Camera microphones also record background noise with other background sound. This noisy data has been filtered out using the Fourier transform. When peak locations for each footstep sound are determined, and a Full Width at Half Maxima is computed for each peak, the number of frames within this width are counted, these frames are verify the presence of a moving object.

Moving object detection is an important and basic step in any video surveillance application. Detected object has used an input to higher level tasks such as tracking, event detection, and behavioral analysis. There are several challenges in moving object detection such as illumination changes, occlusion, shadow, bootstrap, dynamic background etc. Detection of moving object having similar color as background structure is called as camouflage, which is another big challenge. The method of detecting camouflage moving object is known as Decamouflaging. In recent years, the researchers are attracted towards the camouflage moving object detection due to its application in the military. In this review paper, different techniques using various features for camouflaged moving obj ect detection were studied.