Timely and high-density air quality monitoring is essential for the development of future smart cities. The images captured from widely deployed stationary-cameras can be transferred quickly via the ...Internet of Things (IoT) to facilitate ambient pollution estimation anytime anywhere. Image-based air pollution estimation is normally formulated as a supervised learning problem, relying on an extended number of image samples. However, individual stationary-cameras can offer only very limited samples and scenes, while locally trained estimation models can easily overfit. A global method was proposed to address this challenge. The global model was trained via images captured from different cameras. However, such a model is less effective in extracting local features from scenes. A personalized method is therefore proposed to improve not only the generalization of the estimation model but also to preserve the local characteristics of individual cameras. Our personalized method consists of a two-stage architecture: 1) images from different cameras are used to train the global estimation model to avoid overfitting due to fixed scenes and small sample size and 2) the global model is further refined by images captured from individual cameras separately for adapting local characteristics. To evaluate our proposed personalized method, a large data set was constructed, based on stationary-camera-taken images captured in Hong Kong, consisting of different pollution measurements, including PM2.5, PM10, NO2, and O3. As compared to the local model, our proposed personalized model has reduced average MAE by 5.68% and average SMAPE by 6.82%, and improved average <inline-formula> <tex-math notation="LaTeX">r </tex-math></inline-formula> by 4.69%.
In this study, the authors present a new approach to detect fire flame by processing and analysing the stationary camera videos. For a fire detection system, it is desired to be sensitive and ...reliable. The proposed method improves not only the sensitivity but also the reliability through reducing the susceptibility to false alarms. The proposed approach based on multi-feature, i.e. chromatic features, dynamic features, texture features, and contour features, can both improve the sensitivity and reliability in fire detection. In their approach, the authors adopt a novel algorithm to extract the moving region and analyse the frequency of flickers. Experimental results show that the proposed method can run in real-time and performs favourably against the state-of-the-art methods with higher accuracy in fire videos, lower false alarm rates in non-fire videos and faster response time.
We quantify the percentage of sea surface covered by whitecaps from images taken by a non-stationary camera mounted on a moored buoy using an Adaptive Thresholding Segmentation (ATS) method and an ...Iterative Between Class Variance (IBCV) approach. In the ATS algorithm, the optimal value for the threshold is determined as the last inflection point of the smoothed cumulative histogram of the scene. This makes the method more effective in finding the optimal value of the threshold and reduces the computational efforts compared to the conventional Automated Whitecap Extraction (AWE) technique. In the IBCV method, the optimum criterion for determining the value of the threshold corresponds to the measure of separability between the segmented water and whitecap pixels. In our experiments, the fraction of each image covered by the whitecap is determined using the aforementioned dynamical thresholding techniques for images taken under complex forcing and lighting conditions. Comparisons between different techniques suggest the effectiveness of the proposed methodologies, in particular the ATS algorithm to separate the whitecap features from the darker water pixels.
•Adaptive Thresholding Segmentation using histogram information for detection of whitecap areas.•Iterative between class variance method to separate whitecap pixels from the darker water pixels.•Estimates of whitecap fraction from threshold-based automated techniques.
Summary
Non‐rigid moving multiple objects detection and tracking play an important role in intelligent video surveillance system, autonomous navigation, and activity analysis. Closed Circuit ...Television (CCTV) systems are deployed in numerous areas such as airports, traffic intersections, underground stations, mass events, mall, schools, and organisations for security and public surveillance. Although these cameras record continuous video 24x7, it is a human constraint to manually monitor all events such as crime, terrorism, hideous, suspicious activities, the positioning of the vehicle, and fire recorded from a number of cameras. Moreover, problems like dynamic background, the creation of ghost, sensor noise, varying illumination, and colour and compression artefacts affect effective detection of multiple moving objects. This study presents an effective approach named as enhanced Fractal Texture Analysis with KNN classifier (FTAKC) for tracking and detection of multiple objects from a video sequence. The proposed approach comprises three main phases, namely, detection of moving object, tracking of the object (enhanced Fractal Texture Analysis), and behaviour analysis for activity recognition (KNN classifier). The image feature has been extracted based on colour, texture, and geometry were used to identify and track multiple objects in video frames, and Problem domain knowledge rules were applied to distinguish normal or anomalous activities as well as behaviours. Edge detection algorithm (Intersection over Union (IoU) threshold to determine possible edge connections) was applied toward enhancing the illumination variation by multi‐block Local Binary Pattern (LBP) temporal‐analysis to do the segmentation. Finally, the efficiency and effectiveness of the proposed approach has been estimated based on the measure of average PSNR, precision, recall, f‐measure, accuracy, and execution time. The Laboratory for Image and Media Understanding (LIMU) dataset has been utilised toward illustrating the robustness of the proposed approach. Furthermore, it evaluated the performance based on the measure of precision, recall, and F‐measure metrics. It has been tentatively demonstrated that the proposed approach is suitable for recognizing multiple moving object with detection accuracy up to 93.56%. The simulated results show that suggested approach is robust, flexible, as well as able to outperform the traditional methods than the present object detection method.
This study presents the Segmented Colour Feature Extreme Learning Machine (SCF-ELM). The SCF-ELM is inspired by the Extreme Learning Machine (ELM) which is known for its rapid training and inference ...times. The ELM is therefore an ideal candidate for an ensemble learning algorithm. The Colour Feature Extreme Learning Machine (CF-ELM) is used in this study due to its additional ability to extract colour image features. The SCF-ELM is an ensemble learner that utilizes feature mapping via k-means clustering, a decision matrix and majority voting. It has been evaluated on a range of challenging agricultural object classification scenarios including weed, livestock and machinery detection. SCF-ELM model performance results were excellent both in terms of detection, 90 to 99% accuracy, and also inference times, around 0.01(s) per image. The SCF-ELM was able to compete or improve upon established algorithms in its class, indicating its potential for remote computing applications in agriculture.
•We propose to apply low-rank and sparse decomposition (LRSD) for video compression.•We propose an incremental LRSD (ILRSD) which facilities large-scale video processing.•Our method achieves better ...coding efficiency compared to the state-of-the-art.
Videos captured by stationary cameras are usually with a static or gradually changed background. Existing schemes are not able to globally exploit the strong background temporal redundancy. In this paper, motivated by the recent advance on low-rank and sparse decomposition (LRSD), we propose to apply it for the compression of videos captured by fixed cameras. In particular, the LRSD is employed to decompose the input video into the low-rank component, representing the background, and the sparse component, representing the moving objects, which are encoded by different methods. Moreover, we further propose an incremental LRSD (ILRSD) algorithm to reduce the large memory requirement and high computational complexity of the existing LRSD algorithm, which facilitates the process of large-scale video sequences without much performance loss. Experimental results show that the proposed coding scheme can significantly improve the existing standard codecs, H.264/AVC and HEVC, and outperform the state-of-the-art background modeling based coding schemes.
The existing tracking and recognition methods concentrate mainly on single-class targets; however, systems for traffic management or intelligent transport often require multi-class target tracking ...and recognition in real time. This study proposes an effective multi-class moving target recognition method that is based on Gaussian mixture part-based model, which accurately locates objects of interest and recognises their corresponding categories. The method is multi-threaded and combines soft clustering approach with multiple mixture part based models to provide stable multi-class target tracking and recognition in video sequences. The highlight of the method is its ability to recognise multi-class moving targets and to count their numbers in the video sequence captured by a stationary camera with fixed focal length. Another contribution of this study is that an extended part based model is developed for object recognition in real-world environments, which can improve the overall system performance, lower time costs, and better meet the actual demand of a video system. Experimental results show that the proposed method is viable in real-time multi-class moving target tracking and recognition.
Object detection and tracking is an important task within the field of computer vision, because of its promising application in many areas, such as video surveillance. The need for automated video ...analysis has generated a great deal of interest in the area of motion tracking. A new technique is proposed for online object tracking-by-detection capable of achieving high detection and tracking rates, using a stationary camera, in a particle filtering framework. The fundamental innovation is that the detection technique integrates the local binary pattern texture feature, the red green blue (RGB) colour feature and the Sobel edge feature, using ‘Choquet’ fuzzy integral to avoid uncertainty in the classification. This is performed by extracting the colour and edge grey scale confidence maps and introducing the texture confidence map. Then, the tracking technique makes use of the continuous confidence detectors, extracted from those confidence maps, along with another three introduced classifier confidence maps, extracted from an online boosting classifier. Finally, both the confidence detectors and the classifier maps are integrated in the particle filtering framework, using the Choquet integral. Experimental results for both indoor and outdoor dataset sequences confirmed the robustness of the proposed technique against illumination variation and scene motion.
We use foreground pixels as indicators of the presence of people to estimate the total number of people in an image sequence; the proposed method has low processing cost and can output the number in ...real time. The number of people is computed from the visual hull for each pixel, based on the camera parameters of a stationary camera, and a simple geometric model for each person. The proposed method uses a genetic algorithm to automatically adjust the camera parameters, in the framework of calculating the total number of people. We evaluate the performance of the method with computer simulations and demonstrate its effectiveness on real images.