Image recognition technology is an important field of artificial intelligence.Combined with the development of machine learning technology in recent years,it has great researches value and commercial value.As a matter...Image recognition technology is an important field of artificial intelligence.Combined with the development of machine learning technology in recent years,it has great researches value and commercial value.As a matter of fact,a single recognition function can no longer meet people’s needs,and accurate image prediction is the trend that people pursue.This paper is based on Long Short-Term Memory(LSTM)and Deep Convolution Generative Adversarial Networks(DCGAN),studies and implements a prediction model by using radar image data.We adopt a stack cascading strategy in designing network connection which can control of parameter convergence better.This new method enables effective learning of image features and makes predictive models to have greater generalization capabilities.Experiments demonstrate that our network model is more robust and efficient in terms of timing prediction than 3DCNN and traditional ConvLSTM.The sequential image prediction model architecture proposed in this paper is theoretically applicable to all sequential images.展开更多
Solar forecasting using ground-based sky image offers a promising approach to reduce uncertainty in photovoltaic(PV)power generation.However,existing methods often rely on deterministic predictions that lack diversity...Solar forecasting using ground-based sky image offers a promising approach to reduce uncertainty in photovoltaic(PV)power generation.However,existing methods often rely on deterministic predictions that lack diversity,making it difficult to capture the inherently stochastic nature of cloud movement.To address this limitation,we propose a new two-stage probabilistic forecasting framework.In the first stage,we introduce I-GPT,a multiscale physics-constrained generative model for stochastic sky image prediction.Given a sequence of past sky images,I-GPT uses a Transformer-based VQ-VAE.It also incorporates multi-scale physics-informed recurrent units(Multi-scale PhyCell)and dynamically weighted fuses physical and appearance features.This approach enables the generation of multiple plausible future sky images with realistic and coherent cloud motion.In the second stage,these predicted sky images are fed into an Image-to-Power U-Net(IP-U-Net)to produce 15-min-ahead probabilistic PV power forecasts.In experiments using our dataset,the proposed approach significantly outperforms deterministic,other stochastic,multimodal,and smart persistence baselines models,achieving a superior reliability–sharpness trade-off.It attains a Continuous Ranked Probability Score(CRPS)of 2.912 kW and a Winkler Score(WS)of 33.103 kW on the test set and CRPS of 2.073 kW and WS of 22.202 kW on the validation set.Translating to 35.9%and 42.78%improvement in predictive skill over the smart persistence model.Notably,our method excels during rapidly changing cloud-cover conditions.By enhancing both the accuracy and robustness of short-term PV forecasting,the framework provides tangible benefits for Virtual Power Plant(VPP)operation,supporting more reliable scheduling,grid stability,and risk-aware energy management.展开更多
A predictive search algorithm to estimate the size and direction of displacement vectors was presented.The algorithm decreased the time of calculating the displacement of each pixel.In addition,the updating reference ...A predictive search algorithm to estimate the size and direction of displacement vectors was presented.The algorithm decreased the time of calculating the displacement of each pixel.In addition,the updating reference image scheme was used to update the reference image and to decrease the computation time when the displacement was larger than a certain number.In this way,the search range and computational complexity were cut down,and less EMS memory was occupied.The capability of proposed search algorithm was then verified by the results of both computer simulation and experiments.The results showed that the algorithm could improve the efficiency of correlation method and satisfy the accuracy requirement for practical displacement measuring.展开更多
A method is presented to convert any display screen into a touchscreen by using a pair of cameras. Most state of art touchscreens make use of special touch-sensitive hardware or depend on infrared sensors in various c...A method is presented to convert any display screen into a touchscreen by using a pair of cameras. Most state of art touchscreens make use of special touch-sensitive hardware or depend on infrared sensors in various configurations. We describe a novel computer-vision-based method that can robustly identify fingertips and detect touch with a precision of a few millimeters above the screen. In our system, the two cameras capture the display screen image simultaneously. Users can interact with a computer by the fingertip on the display screen. We have two important contributions： first, we develop a simple and robust hand detection method based on predicted images. Second, we determine whether a physical touch takes places by the homography of the two cameras. In this system, the appearance of the display screen in camera images is inherently predictable from the computer output images. Therefore, we can compute the predicted images and extract human hand precisely by simply subtracting the predicted images from captured images.展开更多
基金This work was supported in part by the Open Research Project of State Key Laboratory of Novel Software Technology under Grant KFKT2018B23the Priority Academic Program Development of Jiangsu Higher Education Institutions,and the Open Project Program of the State Key Lab of CAD\&CG(Grant No.A1916),Zhejiang University.
文摘Image recognition technology is an important field of artificial intelligence.Combined with the development of machine learning technology in recent years,it has great researches value and commercial value.As a matter of fact,a single recognition function can no longer meet people’s needs,and accurate image prediction is the trend that people pursue.This paper is based on Long Short-Term Memory(LSTM)and Deep Convolution Generative Adversarial Networks(DCGAN),studies and implements a prediction model by using radar image data.We adopt a stack cascading strategy in designing network connection which can control of parameter convergence better.This new method enables effective learning of image features and makes predictive models to have greater generalization capabilities.Experiments demonstrate that our network model is more robust and efficient in terms of timing prediction than 3DCNN and traditional ConvLSTM.The sequential image prediction model architecture proposed in this paper is theoretically applicable to all sequential images.
基金supported by the“Regional Innovation Strategy(RIS)”through the National Research Foundation of Korea(NRF)funded by the Ministry of Education(MOE)(2021RIS002)the Technology Development Program(RS-2025-02312851)funded by the Ministry of SMEs and Startups(MSS,Republic of Korea).
文摘Solar forecasting using ground-based sky image offers a promising approach to reduce uncertainty in photovoltaic(PV)power generation.However,existing methods often rely on deterministic predictions that lack diversity,making it difficult to capture the inherently stochastic nature of cloud movement.To address this limitation,we propose a new two-stage probabilistic forecasting framework.In the first stage,we introduce I-GPT,a multiscale physics-constrained generative model for stochastic sky image prediction.Given a sequence of past sky images,I-GPT uses a Transformer-based VQ-VAE.It also incorporates multi-scale physics-informed recurrent units(Multi-scale PhyCell)and dynamically weighted fuses physical and appearance features.This approach enables the generation of multiple plausible future sky images with realistic and coherent cloud motion.In the second stage,these predicted sky images are fed into an Image-to-Power U-Net(IP-U-Net)to produce 15-min-ahead probabilistic PV power forecasts.In experiments using our dataset,the proposed approach significantly outperforms deterministic,other stochastic,multimodal,and smart persistence baselines models,achieving a superior reliability–sharpness trade-off.It attains a Continuous Ranked Probability Score(CRPS)of 2.912 kW and a Winkler Score(WS)of 33.103 kW on the test set and CRPS of 2.073 kW and WS of 22.202 kW on the validation set.Translating to 35.9%and 42.78%improvement in predictive skill over the smart persistence model.Notably,our method excels during rapidly changing cloud-cover conditions.By enhancing both the accuracy and robustness of short-term PV forecasting,the framework provides tangible benefits for Virtual Power Plant(VPP)operation,supporting more reliable scheduling,grid stability,and risk-aware energy management.
文摘A predictive search algorithm to estimate the size and direction of displacement vectors was presented.The algorithm decreased the time of calculating the displacement of each pixel.In addition,the updating reference image scheme was used to update the reference image and to decrease the computation time when the displacement was larger than a certain number.In this way,the search range and computational complexity were cut down,and less EMS memory was occupied.The capability of proposed search algorithm was then verified by the results of both computer simulation and experiments.The results showed that the algorithm could improve the efficiency of correlation method and satisfy the accuracy requirement for practical displacement measuring.
文摘A method is presented to convert any display screen into a touchscreen by using a pair of cameras. Most state of art touchscreens make use of special touch-sensitive hardware or depend on infrared sensors in various configurations. We describe a novel computer-vision-based method that can robustly identify fingertips and detect touch with a precision of a few millimeters above the screen. In our system, the two cameras capture the display screen image simultaneously. Users can interact with a computer by the fingertip on the display screen. We have two important contributions： first, we develop a simple and robust hand detection method based on predicted images. Second, we determine whether a physical touch takes places by the homography of the two cameras. In this system, the appearance of the display screen in camera images is inherently predictable from the computer output images. Therefore, we can compute the predicted images and extract human hand precisely by simply subtracting the predicted images from captured images.
