414results about How to "Solve occlusion" patented technology

The invention relates to a face feature extraction method based on a face feature point shape drive depth model. The method comprises the steps that the face feature point shape drive depth model is set up, N depth convolution neural networks are utilized for extracting features of N face regions divided according to the positions of face feature points to obtain the discrimination feature and the attributive feature of each region, and then all the discrimination features and the attributive features are fused to obtain features higher in descriptive ability. According to the face feature extraction method based on the face feature point shape drive depth model, the problem of robustness under change conditions of illumination, angles, expressions, shielding and the like can be well solved, and the recognition rate of face recognition under these conditions is increased.

The invention discloses an automobile A pillar perspective vehicle-mounted display device which reflects dynamic changes of driver sight angles. Cameras are arranged both inside an automobile and outside the automobile, the cameras outside the automobile are used for collecting surrounding sight information of a road ahead, and the cameras inside the automobile are used for obtaining driver vision information; according to driver sight concerning angles, images concerned by a driver are cut from images obtained by the cameras outside the automobile; flexible liquid crystal display screens are arranged on the inner side of an A pillar in an attaching mode according to the radian of the A pillar to display images; the images displayed on the flexible liquid crystal display screens inside the automobile are changed in real time according to driver sight angle change information collected by the cameras inside the automobile; the angle and the size of a driving scene image, collected by the cameras outside the automobile, at the position of a blind region are changed in real time, and an A pillar display system can be adjusted in a self-adaption mode according to sight needing regions detected through sight tracking. Therefore, external road traffic environment information which is more complete, more accurate and good in driver acceptability is provided for the driver, and the safety and the convenience of automobile driving are improved.

The invention belongs to the field of security and protection monitoring and relates to a multi-level characteristic data association-based method for detecting and tracking multiple targets in a monitored scene by adopting a multi-target visual tracking algorithm. The method comprises the following steps: acquiring target objects, and extracting targets from a video scene background to serve as to-be-tracked objects; establishing target models, extracting global features and local features of the targets to serve as features of described target object models, and removing the targets with very low matching degree by filtering; performing association calculation by taking target similarity, texture similarity and local color similarity as the basis, and performing a matching strategy from two levels of the global features and the local features; processing target occlusion by adopting an occlusion judging mechanism; and obtaining target trajectories which are longer and comparatively accurate according to an association algorithm of subsequent iteration. According to the method provided by the invention, not only is the accuracy of target tracking improved but also the occlusion issue among the targets is effectively solved. Compared with an association tracking algorithm only using overall feature modeling, the method provided by the invention is higher in accuracy.

The invention discloses a passenger flow statistics method based on a computer visual perception and system thereof. A video image collecting device can be placed at top end of an entrance or an exit of passages such as airport, subway, railway station, bus station, exhibition and so on to collect video images of passenger flows. A processor processes the video images collected by a computer visual perception algorithm to obtain a characteristic point of the image and to obtain a characteristic point trace by processing continuous image sequences within a period of time. Restricting the characteristic trace in time-space obtains smooth and continuous trace and clusters obtained characteristic traces to obtain number of passenger flows so as to obtain real time and accurate passenger flow information. The invention combines visual device and algorithm of computer to count number of the passenger flow accurately under a compact passenger flow circumstance. The obtained passenger flow information can be used as ground of safety monitor, operation management and line layout.

The invention discloses an indoor visible light positioning system. The indoor positioning technology is completed jointly by combination of an LED lamp, a hand-held receiving terminal, and a positioning algorithm. Position information is modulated on a carrier wave distributed to the LED lamp and reverse fast Fourier transform is carried out; a signal is sent into a drive circuit and then the signal is transmitted to an air channel in a visible light mode by free space. A receiving terminal carries out light convergence by an optical lens; the signal is converted into an electric signal; fast Fourier transform is carried out and digital demodulation is carried out; and according to information of three detected strongest powers, a distance between the receiving terminal and the LED lamp is estimated by a lambertian model. According to position coordinate information of a carried transmitting terminal and estimated distance information, a receiving coordinate is calculated finally and positioning is carried out. According to the invention, communication reliability is enhanced; the positioning precision is improved; and the interference of communication between LED cells is effectively resisted.

A pedestrian re-identification method, system, electronic device and computer-readable storage medium are disclosed. The method includes: acquiring a video set and determining a target pedestrian image; Extracting target features of the target pedestrian image, and inputting the target features into the NSN to extract an image to be recognized of each frame of the image in the video set; Generating an optical flow map of an image to be recognized of adjacent frames by using an MN network; All the images to be identified and all the optical flow images are input into the LSTM network to obtaina fusion picture fusing multi-frame image information. The global feature map and partial attention map of the fusion image are extracted from the trained human feature extraction network, and each part of the attention map and the global feature map are fused into a fused partial attention feature map respectively. The fusion feature vectors of each part of the attention fusion feature map are formed by the global average pool, and all the fusion feature vectors are connected to the global feature vectors, which improves the recognition accuracy of pedestrian re-recognition.

The invention provides a robot indoor environment exploration,obstacle avoidance and target tracking method based on a ROS. Based on a grid map established through laser radar information,combined with local map deduction and global boundary searching,an independent exploration strategy is designed so that a moving robot can be prevented from getting into a local exploration infinite loop,and whole-indoor-environment exploration can be completely guaranteed. According to the method,real-time tracking nodes and shielding tracking nodes are designed with an improved Kalman filtering method and aMeanShift method in the ROS,the real-time problem and the complete-shielding problem of target tracking of the moving robot are solved,the computing speed of the system is increased,target search time is shortened,and the real-time requirement of tracking is met; when a target is completely shielded,the shielding tracking nodes forecast and track the target according to previous state information,and after shielding is ended,a target is newly locked,tracking modes are automatically switched,and the target is tracked through the real-time tracking mode.

The invention discloses a dual-band co-aperture flat array antenna which is applied to multiband homing heads and intelligent ammunition guidance systems. The dual-band co-aperture flat array antenna comprises a single-ridge waveguide slot array surface, a substrate integrated waveguide slot array surface, single-ridge waveguide slot array feed networks and a substrate integrated waveguide slot array feed network. The single-ridge waveguide radiation slot array surface comprises a plurality of rows of parallel single-ridge waveguide radiation slot line arrays; a substrate integrated waveguide radiation slot array comprises a plurality of rows of parallel substrate integrated waveguide radiation slot line arrays; an antenna array surface is divided into four quadrants; the feed networks respectively feed array surfaces of the four quadrants so as to have the condition of realizing a single-pulse function; each quadrant has two sets of feed networks for respectively feeding the single-ridge waveguide slot line arrays and the substrate integrated waveguide slot line arrays; the single-ridge waveguide slot array surface works in an X-band; the substrate integrated waveguide slot array surface works in a Ka-band; arrays of two forms have a common aperture and do not mutually affect; and the problem of electromagnetic wave blocking caused by non-coplanar array surfaces is solved.

The invention discloses a motion capture method based on inertia and optical measurement fusion, and relates to the field of a sensing technology and computer vision, in particular to capture and analysis of motion processes of one or more objects. An inertia measurement technology and an optical capture technology are combined. The inertia measurement technology has the characteristics that the inertia measurement technology is insusceptible to the external environment, but problems of error accumulation, data drifting, poorer dynamic tracking performance and the like exist during measurement; the optical motion capture technology is high in accuracy but is limited by the environment such as illumination, colors and the like easily and interfered by shielding. The motion capture system and method integrate advantages of the inertia measurement technology and the optical capture technology, are low in cost, high in accuracy, small in error and insusceptible to the environment and are not interfered by shielding.

The invention discloses a method and a system for determining binocular scene flow based on semantic segmentation. Firstly, the scene in binocular image is semantically segmented, and the semantic optical flow is calculated by adding semantic segmentation label information, then the disparity information is calculated by semi-global matching algorithm, and then the motion parameters of countless facet regions are calculated and optimized by combining the semantic optical flow and disparity information fitting. In the process of optimization, the initial scene flow is obtained by superpixel segmentation, and then the motion of the superpixel blocks in the semantic tag is optimized, so the motion of the superpixel blocks in the semantic tag is consistent and the edge information of the moving object is protected well. Semantic information is added into the optical flow information, so that the edge of the object is protected, and the reasoning process of the occlusion problem is greatlysimplified. In addition, the motion inference at the semantic tag level makes the scene flow of pixels on the surface of the same moving object approximately consistent, and ultimately the goal of optimizing the scene flow is achieved.

In order to solve the problems that the prior transformation method on video frame rate cannot solve moving object sheltering and irregular motion well, the invention provides a video frame rate upconversion method, which comprises the following steps: ranking a video sequence to be processed as a film mode video sequence or a non-film mode video sequence; carrying out frame selection of the film mode video sequence; carrying out motion estimation and motion compensation processing of the film mode video sequence processed in the previous step; and carrying out processing such as low-frequency frame insertion or repeated frame insertion for the non-film mode video sequence. The video frame rate upconversion method can be used for frame rate conversion of a video sequence.

The invention discloses a monocular light field image unsupervised depth estimation method based on a convolutional neural network. According to the method, the disclosed large-scale light field imagedata set is firstly used as a training set, and samples of the training set tend to be balanced through data enhancement and data expansion; an improved ResNet50 network model is constructed; an encoder and a decoder are used for extracting high-level and low-level features of a model respectively, results of the encoder and the decoder are fused through a dense difference structure, meanwhile, asuper-resolution shielding detection network is additionally constructed, and the shielding problem between all visual angles can be accurately predicted through deep learning; the objective functionbased on the light field image depth estimation task is a multi-loss function, the preprocessed image is trained through a pre-defined network model, and finally generalization evaluation is carriedout on the network model on a test set. According to the method, the preprocessing effect on the light field image of the complex scene is obvious, and the effect of more accurate light field image unsupervised depth estimation is achieved.

The invention discloses an adaptive anti-occlusion infrared target tracking method based on multi-layer depth feature fusion. Firstly, a series of multi-layer depth feature maps of the same size and different levels are obtained; the multi-layer depth feature map is then converted from a time domain to a frequency domain according to correlation filtering, a filter is trained and the response mapis calculated according to the fast Fourier transformation, and then the multi-layer depth feature map is merged and dimensionally reduced according to the weighted fusion of the intra-layer features,so that the feature response map of different levels is constructed and the maximum correlation response value is obtained, which is the estimated position of the target; at last, that dense featuresof the target are extracted, and the response confidence of the target center position estimated by the depth convolution feature is obtained according to the maximum response value of the feature obtained by the correlation filtering; when the response confidence of the target center position is less than the re-detection threshold T0, the obtained target estimated position is evaluated by on-line target re-detection and the position of the target is adaptively updated according to the evaluation result.

The invention provides a single-axis rotation angle vision measurement method based on a fixed camera and a single target. The method includes the four steps of firstly, calibrating internal parameters and external parameters of the camera; secondly, obtaining the initial values of the external parameters related to pivoting movement; thirdly, conducting nonlinear optimization on the internal parameter pin and the external parameter pout; fourthly, measuring the rotation angle through the calibrated internal parameters and the calibrated external parameters. The method is a method for measuring the rotation angle of the pivoting movement through the monocular camera, and solves the problem that the target is hard to install in an existing angle vision measurement method; meanwhile, the method is suitable for the conditions with blockage as well and is quite high in practicability.

The present invention discloses a spatial position and attitude determination system based on a virtual reality (VR) technology and a method therefor. The spatial position and attitude determination system mainly comprises a VR helmet, two base stations, two VR handles and a computer. Wireless communication is performed between the two VR handles and the VR helmet, between the base stations and the VR helmet, and between the helmet and the computer all by means of a Bluetooth module. By application of the spatial position and attitude determination system and the method therefor provided by the present invention, the technical defect that inherent measurement errors exist in the VR system spatial position and attitude determination system based on an inertial sensor and will be accumulated along with an integration process, which causes position drift so that continuous and accurate spatial position measurement data cannot be provided can be effectively overcome.

A method and an apparatus are used for detecting a continuous road partition with a height, the method comprising: a step of obtaining a disparity map including a part of the continuous road partition, a corresponding U-disparity map and a corresponding V-disparity map; and a step of detecting the continuous road partition based on the U-disparity map corresponding to a slice parallel to the ground. According to the method and the apparatus for detecting the continuous road partition with a height according to the embodiment of the present invention, the disparity map, and the U-disparity map or the V-disparity map are used, therefore problems of environmental change and geometric deformation can be solved successfully.

The invention discloses a joint reasoning-based video multi-target tracking method in the technical field of video processing. The method comprises the following steps: firstly, reading a frame of image of a video file, carrying out image rasterization processing, and then calibrating a candidate position of a target by adopting an online detector and a KLT tracking algorithm as a tracker, respectively screening and then integrating the results; secondly, carrying out quantitative grading on the obtained candidate position result; finally, describing the target tracking condition by using a joint function, and taking the optimal solution based on the joint function as the position of the target in the frame, so as to achieve target tracking. By adopting the joint reasoning-based video multi-target tracking method, the processing of the treatment method combined with a detection tracking algorithm and multi-target relationship in the tracking technology under multi-target tracking can be achieved, and the multi-target relationship is described by using the joint function, so that not only is the result fusion of detection and tracking achieved, but also the relationship between the targets is integrated according to overall situation, so that a global optimal resolution is obtained.

The invention discloses a cable rope bearing bridge deck vehicle load distribution real-time detection method. The cable rope bearing bridge deck vehicle load distribution real-time detection method comprises the steps of shooting a bridge deck image; conducting perspective correction processing and enhancement processing on the image so as to obtain a bridge deck image subjected to image enhancement and adopting an edge information based detection method to judge a bridge deck vehicle; tracking a vehicle image and correcting a deformed structure of the vehicle in the acquired image; using the bridge deck as an absolute coordinate system, accurately drawing a running track of an automobile tire in each video picture section on the bridge deck, splicing tracks of the same vehicle in different video pictures according to a same track principle to obtain a running track of each vehicle on the bridge deck and achieving vehicle load real-time tracking. The cable rope bearing bridge deck vehicle load distribution real-time detection method is wide in application range, low in cost, capable of obtaining a detection result in real time and small in environmental influence. By means of a perspective correction and image enhancement technology, high-quality earlier-stage images can be obtained. The vehicle blocking problem caused by image acquisition in an oblique state can be well solved by means of a dynamic template matching technology, and the method has a wide application range.

The invention relates to a method for reconstructing a large-scale complex flyover 3D model by using airborne LiDAR data. The method includes the following steps: using inverse iteration mathematic morphological filtering and area information to extract flyover point cloud from airborne LiDAR data; in accordance with bridge floor connectivity, segmenting the flyover cloud point; adopting a central line vertical line scanning method to segment and extract a communicated bridge surface which has no bifurcation or intersection structures and has a width consistent width, acquiring a "constitutional unit"; an end place of the constitutional unit central line serving as a circular buffer area, detecting a shielding structure of the flyover through determination of point cloud height information within the buffer area; through steps of constitutional unit matching, 2D central line fitting and 3D curved surface fitting, acquiring a complete bridge surface 3D central line; and reconstructing a flyover 3D model in combination with bridge surface width information. Practice has proved that the method of the invention can effectively reconstruct the large-scale complex flyover 3D model, addresses shielding problem of the flyover point cloud, and the reconstructed 3D model has a higher correctness and completeness.

The invention discloses an image processing video tracking technology. The technology specifically comprises tracking the movement of drills in a punching video, measuring and analyzing tracking paths to automatically calculate the number of drilling rods, and calculating the depth of drilling holes according to the number of the drilling rods. The measuring error caused by human factors can be avoided, the data can be processed in real time or after the event; the measuring security is excellent, and major hidden danger can be solved. A method for measuring the coal mine underground drilling depth based on an image processing technology is provided, and the technology for automatically tracking and identifying objects in the video with low-quality in the underground complex environment is provided; according to the method, the problems that the underground movement objects are complex, and underground workers block the drilling hole video can be solved; the complex light environment generated by underground mine lamps and head lamps of the workers can be processed; the tracking result automatic quantitative analysis and counting judging result precision is high, and the real-time performance of a system is high.

The invention provides a target tracking method and device. The method includes the steps of: collecting a depth image and a color image of a current video image; obtaining at least two target regions in the depth image according to the depth image; performing face detection on the color image corresponding to each target region, and obtaining a face region corresponding to each target region; and when the distance between a first face region corresponding to a first target region and a second face region corresponding to a second target region is smaller than a preset threshold value, using a coring coherent filter to track the first face region and the second face region. The target tracking method realizes accurate tracking of multiple targets when there is the target shadow interference in the multiple target tracking and the shielding interference during target interaction.

The invention relates to an unmanned aerial vehicle with a 360 degree panoramic aerial photographing and video recording effect and a rack frame. Omnibearing panoramic aerial photographing is achieved. In the ascending and descending state of the unmanned aerial vehicle, image information of the unmanned aerial vehicle in the horizontal direction is obtained through cameras, and image information of the unmanned aerial vehicle in the vertical direction is obtained through cameras on a first installation board and a second installation board. The design structure that the unmanned aerial vehicle is surrounded by the six cameras is adopted, the structure is simple, the six cameras are aligned with six faces of a space cube, sub-lens video recording is adopted, in the later manufacturing process, images obtained through the six cameras are spliced to form a spherical picture, a non-dead-angle panoramic spliced picture is constructed, and a vehicle body is effectively prevented from being photographed into the cameras. Meanwhile, the cameras are installed at the ends of the unmanned aerial vehicle, the photographed images cannot be blocked by the vehicle body and can be seamlessly spliced into the panoramic picture in the later period, and the problem that the panoramic aerial photographing cameras are blocked by the unmanned aerial vehicle is solved.

The invention discloses a dynamic display system of a pinyin gliding input keyboard and a display method thereof in the technical field of computer application. The system comprises an input module, a keyboard layout generating module, a keyboard layout display module and a popup window display module, wherein the input module consists of a small-size touch input device and an input unit. Compared with the traditional QWERT keyboard, the keyboard has simple and exquisite arrangement and complete function and greatly reduces the gliding distance during input. Moreover, the problem of shading ascreen by fingers in touch screen input is effectively solved through a small view displayed on the upper part of the keyboard.

The invention provides a video-based vehicle detecting and tracking method and system. The method comprises the steps of inputting a video containing a vehicle; detecting the vehicle in the video by means of a detector, obtaining a detection result of the vehicle, and building an online model of the vehicle according to a first-time detection result of the vehicle; tracking the vehicle in the video, and obtaining a tracking result of the vehicle; utilizing the online model of the vehicle for respectively conducting confidence coefficient calculation on the tracking result and the detecting result of the same fame, and obtaining a final detecting and tracking result according to integration of a confidence coefficient calculation result, the tracking result and the detecting result; utilizing the final detecting and tracking result for updating the online model of the vehicle and the detector through online learning. According to the video-based vehicle detecting and tracking method and system, the problems that the vehicle deforms in the tracking process and other objects shield the vehicle are solved.

The invention relates to a method for detecting heads based on vertically-placed depth cameras. The method comprises the following steps of: a, setting vertically-distributed cameras and an active light source, and acquiring depth images of passersby; b, performing background modeling on the acquired depth images to acquire corresponding foreground images; c, performing simulated random water droplet sprinkling operation on the foreground images; d, taking the water droplets falling to the foreground images as sampling points, comparing the sampling points with adjacent areas thereof until the sampling points move to local minimum value areas and stay in the local minimum value areas, and adjusting the depth numerical values of the local minimum value areas; and e, obtaining a distribution map of the sampling points, comparing the depth numerical values of the local minimum value areas in the distribution map with a preset threshold value, and filtering the local minimum value areas of which the depth numerical values are smaller than the preset threshold value to obtain required head position areas of the passersby, thus realizing the detection of the number of the passersby. The method is convenient to operate, good in robustness and wide in adaptability.

The invention relates to a method for controlling a computer, a touch screen, a projector and other user interfaces through Kinect non-contact gestures. The method comprises the main steps of acquiring actions of double arms of an operator and using a gesture recognition algorithm to recognize gestures, utilizing recognized gestures to achieve interaction between the operator and the computer, the touch screen, projector and other user interfaces. The method relates to six gestures including single click, right-key single click, double clicks, forward rolling, back rolling, zooming in, zooming out and dragging and basically covers the demands for interaction between the operator and the computer, the touch screen, projector and other user interfaces. The method is a novel recognition method, gesture recognition accuracy can be improved, and recognition rapidity can be improved. By the combination of the six gestures, man-machine interaction can be convenient and flexible.

The invention discloses a fast Chinese pinyin input method based on a touch sensitive device and an achieving system thereof. The method includes that a marking tool is used for clicking an initial soft keyboard of a touch screen to input an initial consonant of pinyin; a system pops out a vowel suspension soft keyboard covered around an initial consonant key automatically; the marking tool is used for sliding on the vowel suspension soft keyboard for at most two key positions, and vowels of the pinyin are input; and the marking tool is lifted, and the vowel suspension soft keyboard is hidden automatically. Technical systems of six portions are combined and include (1) a suspension soft keyboard system; (2) seven suspension soft keyboards and key position layouts of the seven suspension soft keyboards; (3) a 'one-key one-pinyin' fast pinyin input method; (4) a 'one-character one-key' sentence input mode; (5) a non-switch stroke assisting technology; (6) a fast part input method. By means of the method and the system, only once key clicking is conducted, simplified pinyin and full pinyin are conducted simultaneously, and efficient inputting can be conducted without changing a pinyin scheme or initial soft keyboard setting. Simultaneously, touch screen size limit is broken through, key false identification rate is low, operation fatigue is small, and high-speed typing can be conducted even if a user does not know the pinyin.

The invention discloses a pedestrian tracking method based on a low-altitude aerial photographing infrared video. Continuous and stable pedestrian tracking is realized through combination of a Lucas-Kanade optical flow method and local area secondary detection. The pedestrian tracking method comprises the steps that 1. an aerial photographing infrared pedestrian support vector machine (SVM) classifier is trained offline; 2. the initial position of a pedestrian target is determined; 3. the pedestrian target is preliminarily tracked by utilizing the LK optical flow method, and the position of the pedestrian target in the next frame is calculated; 4. a search area is set around the predicted position of the pedestrian target; and the infrared pedestrian is secondarily detected in the search area by utilizing the offline trained SVM classifier, and the position of the pedestrian target is updated; and 5. the center of the pedestrian target detected in the search area acts as input coordinates of the LK optical flow method of the next time, and the steps (3)-(5) are repeated. Continuous and stable tracking of the infrared pedestrian target can be realized by the pedestrian tracking method, and the problem of street lamp shielding can also be processed.