Abstract: The invention discloses a person search method based on person re-identification driven localization refinement. On one hand, the region of interest (ROI) conversion module converts an original input image into a small image corresponding to a ROI, and contradiction existing in part of features shared by a person re-identification network and a detection network is avoided; and on the other hand, loss of the person re-identification network can be transmitted back to the detection network in a gradient manner through the ROI conversion module, the supervision of loss of the person re-identification network for the detection bounding box output by the detection network is realized, and the adjusted detection bounding box can effectively remove background interference, contains more useful attribute information and is more suitable for person search, so that the person search accuracy is greatly improved.

Abstract: The invention discloses a person search method based on person re-identification driven localization refinement. On one hand, the region of interest (ROI) conversion module converts an original input image into a small image corresponding to a ROI, and contradiction existing in part of features shared by a person re-identification network and a detection network is avoided; and on the other hand, loss of the person re-identification network can be transmitted back to the detection network in a gradient manner through the ROI conversion module, the supervision of loss of the person re-identification network for the detection bounding box output by the detection network is realized, and the adjusted detection bounding box can effectively remove background interference, contains more useful attribute information and is more suitable for person search, so that the person search accuracy is greatly improved.

Abstract: A spatiotemporal action detection method includes performing object detection on all frames of a sample video to obtain a candidate object set; calculating all interframe optical flow information on the sample video to obtain a motion set; constructing a spatiotemporal convolution-deconvolution network of an attention mechanism and a motion attention mechanism of an additional object; adding both a corresponding sparse variable and a sparse constraint to obtain a network structure S after performing spatiotemporal convolution processing on each time segment of the sample video; training the network structure S with an objective function based on classification loss and loss of the sparse constraint of cross entropy; and calculating an action category and a sparse coefficient corresponding to each time segment of a test sampled video to obtain an object action spatiotemporal location.

Abstract: The present invention discloses a method and system for selecting an image region that facilitates blur kernel estimation, in which the method includes: calculating a relative total variation value of each pixel in a blurred image to obtain a relative total variation mapping image; setting a threshold value to determine whether respective pixel in the image is a boundary pixel or not; then sampling the blurred image and its relative total variation mapping image to obtain a series of image patches; and finally counting the number of boundary pixels in each mapping image patch and selecting out an image region that facilitates blur kernel estimation. According to the method and the system, the problems of excessive dependency on operator experience, low efficiency and the like in the existing region selection methods are effectively solved. The image region that facilitates blur kernel estimation is automatically selected out.

Abstract: A spatiotemporal action detection method includes performing object detection on all frames of a sample video to obtain a candidate object set; calculating all interframe optical flow information on the sample video to obtain a motion set; constructing a spatiotemporal convolution-deconvolution network of an attention mechanism and a motion attention mechanism of an additional object; adding both a corresponding sparse variable and a sparse constraint to obtain a network structure S after performing spatiotemporal convolution processing on each time segment of the sample video; training the network structure S with an objective function based on classification loss and loss of the sparse constraint of cross entropy; and calculating an action category and a sparse coefficient corresponding to each time segment of a test sampled video to obtain an object action spatiotemporal location.

Abstract: The present invention discloses a method and system for selecting an image region that facilitates blur kernel estimation, in which the method includes: calculating a relative total variation value of each pixel in a blurred image to obtain a relative total variation mapping image; setting a threshold value to determine whether respective pixel in the image is a boundary pixel or not; then sampling the blurred image and its relative total variation mapping image to obtain a series of image patches; and finally counting the number of boundary pixels in each mapping image patch and selecting out an image region that facilitates blur kernel estimation. According to the method and the system, the problems of excessive dependency on operator experience, low efficiency and the like in the existing region selection methods are effectively solved. The image region that facilitates blur kernel estimation is automatically selected out.

Abstract: The present invention provides a method for infrared imaging detection and positioning of an underground tubular facility in a plane terrain. Demodulation processing is performed on an original infrared image formed after stratum modulation is generated on the underground tubular facility according to an energy diffusion Gaussian model of the underground tubular facility, so as to obtain a target image of the underground tubular facility.

Abstract: A method for estimating a blur kernel size, the method including: (1) pre-processing a blurred image, to obtain an image, so that a size of the image satisfies an image input size of a multi-class convolutional neural network (CNN); (2) inputting the image into a multi-class CNN with completed training, to obtain a blur-kernel-size probability distribution vector; and (3) comparing each element in the blur-kernel-size probability distribution vector, so that an estimated blur kernel size of the blurred image is the blur kernel size corresponding to a largest element. The invention also provides a system for estimating a blur kernel size. The system includes an image pre-processing module, a training-set synthesizing module, a multi-class CNN module, and a blur kernel size estimation module.

Abstract: A voice control system and method for a multimedia device are provided. The system includes an image sensing module configured to collect a user action image; an image recognizing module configured to determine a type or a status of a control instruction according to the user action image; a voice recognition status managing module configured to activate or wake up the voice recognition program according to a type of a current control instruction; a pickup module configured to collect voice signal; a voice recognizing module configured to recognize the collected voice data to generate a control instruction; and a multimedia function module configured to execute the control instruction to provide a corresponding multimedia function to the user.

Abstract: A method for estimating a blur kernel size, the method including: (1) pre-processing a blurred image, to obtain an image, so that a size of the image satisfies an image input size of a multi-class convolutional neural network (CNN); (2) inputting the image into a multi-class CNN with completed training, to obtain a blur-kernel-size probability distribution vector; and (3) comparing each element in the blur-kernel-size probability distribution vector, so that an estimated blur kernel size of the blurred image is the blur kernel size corresponding to a largest element. The invention also provides a system for estimating a blur kernel size. The system includes an image pre-processing module, a training-set synthesizing module, a multi-class CNN module, and a blur kernel size estimation module.

Abstract: The invention discloses a method for correcting aero-optical thermal radiation noise, comprising steps of: pretreating a degraded image to obtain a multi-scale degraded image group, conducting iteration process of obtaining an optimal solution by using a last scale estimation result as an original value of next scale estimation according to the multi-scale degraded image group, thereby facilitating original-scale bias field estimation, and restoring the degraded image according to the original-scale bias field estimated value thereby obtaining an image after aero-optical thermal radiation noise correction. The invention also discloses a system for correcting aero-optical thermal radiation noise. The invention is capable of solving problems with conventional methods, comprising poor correction effect, high complexity, and incapability in correcting the thermal radiation noise at an image level, and applicable to restoration of an image with aero-optical thermal radiation noise.

Abstract: The invention discloses a method for correcting aero-optical thermal radiation noise, comprising steps of: pretreating a degraded image to obtain a multi-scale degraded image group, conducting iteration process of obtaining an optimal solution by using a last scale estimation result as an original value of next scale estimation according to the multi-scale degraded image group, thereby facilitating original-scale bias field estimation, and restoring the degraded image according to the original-scale bias field estimated value thereby obtaining an image after aero-optical thermal radiation noise correction. The invention also discloses a system for correcting aero-optical thermal radiation noise. The invention is capable of solving problems with conventional methods, comprising poor correction effect, high complexity, and incapability in correcting the thermal radiation noise at an image level, and applicable to restoration of an image with aero-optical thermal radiation noise.

Abstract: An infrared imaging detection and positioning method for an underground building in a planar land surface environment comprises: obtaining an original infrared image g0 formed after stratum modulation is performed on an underground building, and determining a local infrared image g of a general position of the underground building in the original infrared image g0; setting an iteration termination condition, and setting an initial value h0 of a Gaussian thermal diffusion function; using the local infrared image g as an initial target image f0, and performing iteration solution of a thermal expansion function hn and a target image fn by using a maximum likelihood estimation algorithm according to the initial value h0 of the Gaussian thermal diffusion function; and determining whether the iteration termination condition is met, if the iteration termination condition is met, using the target image fn obtained by means of iteration solution this time as a final target image f; and if the iteration termination con

Abstract: The present invention discloses a method for detecting, recognizing, and positioning a zonal underground target in a mountain land environment by detecting a ridge position in the mountain land environment and carrying out energy correction. The method belongs to the interdisciplinary field of pattern recognition, remote sensing technology and terrain analysis. The zonal underground target can cause energy abnormity when the heat field thereof is different from that of a mountain mass, and the heat island effect of the ridge can also cause the energy of the mountain mass to be abnormal. However, the energy abnormity caused by the heat island effect is essentially different from the energy abnormity caused by the zonal underground target in the aspect of mode.

Abstract: A zonal underground structure detection method based on sun shadow compensation is provided, which belongs to the crossing field of remote sensing technology, physical geography and pattern recognition, and is used to carry out compensation processing after a shadow is detected, to improve the identification rate of zonal underground structure detection and reduce the false alarm rate. The present invention comprises steps of acquiring DEM terrain data of a designated area, acquiring an image shadow position by using DEM, a solar altitude angle and a solar azimuth angle, processing and compensating a shadow area, and detecting a zonal underground structure after the shadow area is corrected.

Abstract: The present invention provides a method for infrared imaging detection and positioning of an underground tubular facility in a plane terrain. Demodulation processing is performed on an original infrared image formed after stratum modulation is generated on the underground tubular facility according to an energy diffusion Gaussian model of the underground tubular facility, so as to obtain a target image of the underground tubular facility.

Abstract: A zonal underground structure detection method based on sun shadow compensation is provided, which belongs to the crossing field of remote sensing technology, physical geography and pattern recognition, and is used to carry out compensation processing after a shadow is detected, to improve the identification rate of zonal underground structure detection and reduce the false alarm rate. The present invention comprises steps of acquiring DEM terrain data of a designated area, acquiring an image shadow position by using DEM, a solar altitude angle and a solar azimuth angle, processing and compensating a shadow area, and detecting a zonal underground structure after the shadow area is corrected.

Abstract: The present invention discloses a method for stimulating a temperature field of a mountain mass containing a distributed underground facility under the influence of a seepage effect.

Abstract: The present invention discloses a method for detecting, recognizing, and positioning a zonal underground target in a mountain land environment by detecting a ridge position in the mountain land environment and carrying out energy correction. The method belongs to the interdisciplinary field of pattern recognition, remote sensing technology and terrain analysis. The zonal underground target can cause energy abnormity when the heat field thereof is different from that of a mountain mass, and the heat island effect of the ridge can also cause the energy of the mountain mass to be abnormal. However, the energy abnormity caused by the heat island effect is essentially different from the energy abnormity caused by the zonal underground target in the aspect of mode.

Abstract: An infrared imaging detection and positioning method for an underground building in a planar land surface environment comprises: obtaining an original infrared image g0 formed after stratum modulation is performed on an underground building, and determining a local infrared image g of a general position of the underground building in the original infrared image g0; setting an iteration termination condition, and setting an initial value h0 of a Gaussian thermal diffusion function; using the local infrared image g as an initial target image f0, and performing iteration solution of a thermal expansion function hn and a target image fn by using a maximum likelihood estimation algorithm according to the initial value h0 of the Gaussian thermal diffusion function; and determining whether the iteration termination condition is met, if the iteration termination condition is met, using the target image fn obtained by means of iteration solution this time as a final target image f; and if the iteration termination con