Cílem této práce je vytvořit nástroj umožňující vyhledávání podezřelých osob ve videozáznamu pocházejícího z dohledových kamer. Hledané ...
详细信息
Cílem této práce je vytvořit nástroj umožňující vyhledávání podezřelých osob ve videozáznamu pocházejícího z dohledových kamer. Hledané osoby jsou systému určeny pomocí několika fotografií obličeje. Výstup tvoří informace o výskytu hledaných osob na konkrétních snímcích. Úloha je řešena rozdělením problému na detekci tváře a její následnou identifikaci. Experimenty s existujícími přístupy na vhodných datových sadách zajišťují relevantní porovnání úspěšnosti metod za různých podmínek. Výstupy testů umožňují vybrat vhodné metody a jejich optimální nastavení pro tuto konkrétní úlohu. Práce se zabývá i návrhem vhodné architektury, průzkumem existujících knihoven implementujících zkoumané metody a dalšími způsoby optimalizace výpočtu. Výsledkem je implementace uživatelské aplikace splňující zadané parametry. Její funkce byla otestována na vlastní datové sadě věrně napodobující podmínky reálného provozu.
Atmospheric rivers are important atmospheric features implicated in the global water vapor budget, the cloud distribution, and the associated precipitation. The ARiD (Atmospheric River Detector) code has been develope...
详细信息
Atmospheric rivers are important atmospheric features implicated in the global water vapor budget, the cloud distribution, and the associated precipitation. The ARiD (Atmospheric River Detector) code has been developed to automatically detect atmospheric rivers from water vapor flux and has been applied to the ECMWF ERA5 archive over the period 1980-2020 above the Atlantic Ocean and Europe. A case study of an atmospheric river formed in the East Atlantic on August 2014 that reached France has been detailed using ECMWF ERA5 reanalysis, ground based observation data, and satellite products such as DARDAR, AIRS, GPCP, and GOES. This atmospheric river event presents a strong interaction with an intense upper tropospheric jet stream, which induced stratosphere-troposphere exchanges by tropopause fold. A 1980-2020 climatology of atmospheric rivers over Europe has been presented. The west of France, Iberian Peninsula, and British Islands are the most impacted regions by atmospheric rivers with an occurrence of up to four days per month during the October-April period. Up to 40% of the precipitation observed on the west European coast can be linked to the presence of ARs. No significant trend in the occurrence of the phenomena was found over 1980-2020.
The application of modern intelligent video surveillance system was more and more, and in this paper, an adaptive background model based on color space transformation was constructed and the shadow elimination algorit...
详细信息
The application of modern intelligent video surveillance system was more and more, and in this paper, an adaptive background model based on color space transformation was constructed and the shadow elimination algorithm was optimized. In particular, when the target detection was completed by the RGB space, the detection of the foreground target would be locked to the active contour rectangle. And its real-time tracking was obtained;the target's shadow brightness information was used in the elimination of HSV space. It could be seen from the analysis of the final policy that the algorithm could achieve the goal of video tracking under the condition of changing the size of the target or occlusion. Compared with the traditional algorithm, it had a good advantage.
The research on the three-dimensional counting algorithm of suspension cells is very important for the online counting of cells. Suspension cells are usually observed and counted by taking out cell samples under the m...
详细信息
The research on the three-dimensional counting algorithm of suspension cells is very important for the online counting of cells. Suspension cells are usually observed and counted by taking out cell samples under the microscope. This method cannot realize the online counting of cells. At the same time, there will be weak movement of suspended cells in three-dimensional space, making space counting more difficult. Aiming at the above difficulties, a three-dimensional counting algorithm of suspended cells based on deep learning is proposed. The Faster regions with convolutional neural network (R-CNN) framework is used to identify the target cells, the feature pyramid structure is introduced to reduce the feature loss in the fusion process, and the soft non-maximum suppression algorithm is used to remove the bounding box to improve the detection effect. To reduce the manual labeling workload, an automatic labeling method is proposed. The improved Faster RCNN framework achieves the F1-score of more than 98% after identifying cells, and our method is more accurate than other advanced methods. In terms of three-dimensional counting algorithms, the cells are spatially tracked. Cell matching and correction are carried out by extracting multiple feature information of cells, and adhesion cells are extracted separately for priority processing. The algorithm carried out a tracking experiment on the image sequences of five groups of suspended cells, and the final tracking accuracy of the experiment reached more than 95%. The tracking accuracy is better than other advanced algorithms. Experimental results show that the algorithm achieves high-precision three-dimensional tracking and counting of suspended cells.
Background: In the last few years, some studies have measured heart rate (HR) or heart rate variability (HRV) parameters using a video camera. This technique focuses on the measurement of the small changes in skin col...
详细信息
Background: In the last few years, some studies have measured heart rate (HR) or heart rate variability (HRV) parameters using a video camera. This technique focuses on the measurement of the small changes in skin colour caused by blood perfusion. To date, most of these works have obtained HRV parameters in stationary conditions, and there are practically no studies that obtain these parameters in motion scenarios and by conducting an in-depth statistical analysis. Methods: In this study, a video pulse rate variability (PRV) analysis is conducted by measuring the pulse-to-pulse (PP) intervals in stationary and motion conditions. Firstly, given the importance of the sampling rate in a PRV analysis and the low frame rate of commercial cameras, we carried out an analysis of two models to evaluate their performance in the measurements. We propose a selective tracking method using the Viola-Jones and KLT algorithms, with the aim of carrying out a robust video PRV analysis in stationary and motion conditions. Data and results of the proposed method are contrasted with those reported in the state of the art. Results: The webcam achieved better results in the performance analysis of video cameras. In stationary conditions, high correlation values were obtained in PRV parameters with results above 0.9. The PP time series achieved an RMSE (mean +/- standard deviation) of 19.45 +/- 5.52 ms (1.70 +/- 0.75 bpm). In the motion analysis, most of the PRV parameters also achieved good correlation results, but with lower values as regards stationary conditions. The PP time series presented an RMSE of 21.56 +/- 6.41 ms (1.79 +/- 0.63 bpm). Conclusions: The statistical analysis showed good agreement between the reference system and the proposed method. In stationary conditions, the results of PRV parameters were improved by our method in comparison with data reported in related works. An overall comparative analysis of PRV parameters in motion conditions was more limited due to the lac
The paper focuses on the problem of detecting unmanned aerial vehicles that violate restricted airspace. The main purpose of the research is to develop an algorithm that enables the detection, identification and recog...
详细信息
The paper focuses on the problem of detecting unmanned aerial vehicles that violate restricted airspace. The main purpose of the research is to develop an algorithm that enables the detection, identification and recognition in 3D space of a UAV violating restricted airspace. The proposed method consists of multi-sensory data fusion and is based on conditional complementary filtration and multi-stage clustering. On the basis of the review of the available UAV detection technologies, three sensory systems classified into the groups of passive and active methods are selected. The UAV detection algorithm is developed on the basis of data collected during field tests under real conditions, from three sensors: a radio system, an ADS-B transponder and a radar equipped with four antenna arrays. The efficiency of the proposed solution was tested on the basis of rapid prototyping in the MATLAB simulation environment with the use of data from the real sensory system obtained during controlled UAV flights. The obtained results of UAV detections confirmed the effectiveness of the proposed method and theoretical expectations.
A majority of the existing atmospheric rivers (ARs) detection methods is based on magnitude thresholding on either the integrated water vapor (IWV) or integrated vapor transport (IVT). One disadvantage of such an appr...
详细信息
A majority of the existing atmospheric rivers (ARs) detection methods is based on magnitude thresholding on either the integrated water vapor (IWV) or integrated vapor transport (IVT). One disadvantage of such an approach is that the predetermined threshold does not have the flexibility to adjust to the fast changing conditions where ARs are embedded. To address this issue, a new AR detection method is derived from an image-processing algorithm that makes the detection independent of AR magnitude. In this study, we compare the North Pacific and Atlantic ARs tracked by the new detection method and two widely used magnitude thresholding methods in the present day climate. The results show considerable sensitivities of the detected AR number, shape, intensities and their accounted IVT accumulations to different methods. In many aspects, ARs detected by the new method lie between those from the two magnitude thresholding methods, but stand out with a greater number of AR tracks, longer track durations, and stronger AR-related moisture transport in the AR tracks. North Pacific and North Atlantic ARs identified by the new method account for around 100-120 x 10(3) kg/m/s IVT within the AR track regions, about 50%more than the other two methods. This is primarily due to the fact that the new method captures the strong IVT signals more effectively.
The unknown time-varying parameters of physical laws which govern the dynamics of processes can be obtained from measured input and output signals in a local polynomial approximation procedure. Off-line algorithm is d...
详细信息
The unknown time-varying parameters of physical laws which govern the dynamics of processes can be obtained from measured input and output signals in a local polynomial approximation procedure. Off-line algorithm is developed. The a priori amplitude information about parameters and their derivatives is incorporated to improve the accuracy of the estimation. Mean square errors (MSE) asymptotic with respect to a small sampling period and small window size are presented for the estimates of a fairly arbitrary class of time-varying parameters and their derivatives for the case when the nonlinear dynamics depends on derivatives of the parameters in question.
暂无评论