Song Huihui

Bio: Song Huihui is an academic researcher. The author has contributed to research in topics: Segmentation & Optical flow. The author has an hindex of 2, co-authored 2 publications receiving 5 citations.

Unsupervised video segmentation method integrated with temporal-spatial multi-feature representation

Abstract: The invention discloses an unsupervised video segmentation method integrated with temporal-spatial multi-feature representation. The features of a target are extracted and identified according to themotion information and saliency and color features of the target, and the target is segmented stably and accurately through a Gaussian mixture model. The method includes the following steps: super pixel segmentation; optical flow matching; optimizing the matching result; building a graph model, and calculating the super pixel-level segmentation result; using the segmentation result to train the parameters of a Gaussian mixture model; calculating the pixel-level segmentation result; and getting a final segmentation result according to the super pixel-level segmentation result and the pixel-level segmentation result. Through super pixel segmentation on each frame of image, the computational complexity is greatly reduced. The robustness of segmentation is improved by optimizing the optical flow matching information according to non-local temporal-spatial information. The introduction of the Gaussian mixture model makes up for the large edge matching error in the process of super pixel segmentation. The saliency feature further improves the accuracy and credibility of the segmentation result.

Unsupervised video segmentation method based on non-local space-time characteristic learning

Abstract: The invention discloses an unsupervised video segmentation method based on non-local space-time characteristic learning. The method comprises the following steps of acquiring a video sequence which needs to be segmented; using a superpixel to segment and process the video sequence; using a light stream to carry out previous and next frame information matching; according to adjacent frame information of the video sequence, acquiring a range of a motion target and taking as model initialization input; using global information to optimize a matching result; and establishing a graph model and using a graph segmentation algorithm to solve a segmentation result, and through video segmentation, acquiring output of the motion target. Through carrying out superpixel segmentation on each frame of image in an input video, an operation complexity can be greatly reduced; and non-local space-time information is used to optimize matching information acquired through the light stream so that segmentation robustness can be increased and a noise influence is reduced. Any manual intervention is not needed, and an accurate segmentation result can be acquired completely based on video image information.

Road traffic jam analysis method based on aerial image

Abstract: The invention provides a road traffic jam analysis method based on an aerial image. The method comprises the following steps that: firstly, through an unmanned aerial vehicle, collecting to obtain the aerial image, and transmitting the aerial image to a remote server to be stored; secondly, aiming at the aerial image of a dynamic background to introduce a background compensation technology, reconstructing a static scene in a period of time, adopting a pixel gray level based voting strategy to finish background extraction on the basis, and therefore, utilizing a background subtraction and morphological erosion technology to extract road vehicles; then, adopting an improved mean value drift tracking algorithm to finish the tracking of all vehicles; and finally, combining with a frame rate to calculate time occupancy and average travel speed, taking the time occupancy and the average travel speed as basic indexes for a jam analysis model, establishing a model based on a fuzzy theory, and realizing traffic jam analysis. By use of the method disclosed by the invention, data support can be provided for transport agencies to further carry out traffic jam analysis and a jam dredging strategy, so that a traffic jam phenomenon can be avoided as far as possible, and road traffic resources can be reasonably used to a maximum degree.

Scene moving object segmentation method and system, storage medium and device

Abstract: The invention relates to a scene moving object segmentation method and system, a storage medium and a device. The method comprises the steps of segmenting each foreground target in an image pair composed of two to-be-spliced images with a time difference, and cutting the corresponding to-be-spliced image according to the case segmentation coordinate position of the foreground target obtained through segmentation, so as to obtain the original image region information of the foreground target; determining optical flow field information of the image pair; determining the motion state type of theforeground target; and mapping and marking the instance segmentation result, the instance segmentation coordinate position and the motion state type of each foreground target one by one to obtain a motion target segmentation result. According to the present invention, the motion state type of the foreground target is judged by taking the optical flow field as the motion characteristic informationafter instance segmentation is carried out on the image pair; the motion state of each target is classified while target-level semantic information segmentation is completed, the pixel-by-pixel target-level segmentation of the motion target is achieved, and the anti-noise capability is higher.

Network camera video quality improving method

Abstract: The invention provides a network camera video quality improving method. The method mainly comprises the following steps of (1) obtaining a preliminary video; (2) evaluating the content of the preliminary video, collecting the video data, calibrating and constructing a data set, training by using a convolutional neural network and a full connection network, carrying out feature fusion by using a saliency algorithm to obtain a video quality evaluation model conforming to the sense organ of a person, and judging the video quality obtained by a camera in real time; (3) directly outputting the video with the video quality meeting the requirements, performing the multi-frame video optimization processing based on super pixels on the video which does not meet the requirements, and performing frame reduction optimization; and (4) outputting the video processed in the step (3) through a network. The method has the advantages that the high-accuracy evaluation meeting the human evaluation standards can be carried out on the video quality, more important targets are concerned, the real-time frame reduction optimization is achieved in a super-resolution and multi-frame splicing mode, and the video quality originally outputted by the camera is improved.

Pattern matching method based on texture block matching

Abstract: According to the pattern matching method based on texture block matching, a user can be allowed to realize high-precision and high-consistency ethnic culture pattern segmentation through few interactions, and meanwhile, the quality of vectorization and other ethnic culture pattern digital analysis can be improved. The method comprises the following steps: firstly, finding out all patterns similarto patterns specified by a user by using global block matching; then, detecting a rotation relationship between similar patterns by utilizing local block matching, and further establishing dense correspondence between the similar patterns through constrained block matching; and finally, realizing segmentation collaborative optimization of all similar patterns by utilizing the collaborative optimization model provided by the invention. The method has the advantages of less interaction, high precision, structure protection and the like, can ensure the segmentation consistency between similar patterns, and has a promotion effect on improvement of digital analysis of national culture patterns such as vectorization.

A self-supervised learning model training method and device based on relational reasoning

Abstract: The invention provides a self-supervised learning model training method and device based on relational reasoning. The method comprises obtaining different local observation images corresponding to theimages through different geometric transformation operations; extracting local features corresponding to the corresponding images; fusing the local features to obtain global features of the corresponding images; predicting a corresponding prediction geometric transformation operation between the local feature and the global feature; according to the difference between the prediction geometric transformation operation and the actual geometric transformation operation; constructing a loss function of the learning model; determining target parameters of the learning model through iteration of the loss function; using the prediction geometric transformation operation as a supervision signal to train a learning model. The relationship of the preset auxiliary task is established between the global feature and the local feature, so that the feature obtained by model learning can focus on capture of semantic information of the visual object, the influence of the preset auxiliary task on feature learning is reduced, and migration to the target task is easy.