332 results about "Pedestrian recognition" patented technology

The present invention discloses a surveillance video pedestrian re-recognition method based on ImageNet retrieval. The pedestrian re-recognition problem is transformed into the retrieval problem of an moving target image database so as to utilize the powerful classification ability of an ImageNet hidden layer feature. The method comprises the steps: preprocessing a surveillance video and removing a large amount of irrelevant static background videos from the video; separating out a moving target from a dynamic video frame by adopting a motion compensation frame difference method and forming a pedestrian image database and an organization index table; carrying out alignment of the size and the brightness on an image in the pedestrian image database and a target pedestrian image; training hidden features of the target pedestrian image and the image in the image database by using an ImageNet deep learning network, and performing image retrieving based on cosine distance similarity; and in a time sequence, converging the relevant videos containing recognition results into a video clip reproducing the pedestrian activity trace. The method disclosed by the present invention can better adapt to changes in lighting, perspective, gesture and scale so as to effective improve accuracy and robustness of a pedestrian recognition result in a camera-cross environment.

The invention discloses a pedestrian recognition method based on combination of depth learning and property learning. According to the invention, a convolution neural network containing five implicit strata is constructed. Network training is performed by an anti-convolution method and a concept of property learning is combined. Preferred features obtained from the convolution neural network are input to property classifiers, so that the posterior probability of the property of a sample is obtained. Then by combining with a property class mapping relation, the posterior probability of the class is obtained, so that the class of the sample can be judged. The method is good in detection recognition performance and intrinsic features of an image can be extracted. Besides, since the property has better semantic expression performance than low-stratum features and due to the insensitivity to light and view angles, the algorithm has a good recognition effect.

The invention relates to a pedestrian recognition and tracking method based on an accelerated area Convolutional Neural Network. Firstly, training and testing data set are preprocessed according to the requirements through a robot with an infrared camera to acquire a training dataset and a testing dataset at night, and then, actual target position labeling is conducted on all training and testing photos and is recorded to a sample file; then, the accelerated area Convolutional Neural Network is constructed, the accelerated area Convolutional Neural Network is trained by using the training dataset, and the final probability belonging to a pedestrian area and a bounding box of the area are calculated out from network output by the usage of a non-maximum suppression algorithm; the accuracy of the network is tested by the usage of the testing dataset, and a network model consistent with the requirements is obtained; photos collected by the robot at night are input to an accelerated area Convolutional Neural Network model, and the probability belonging to the pedestrian area and the bounding box of the area are online output by a model in real time. According to the pedestrian detection and tracking method based on the accelerated area Convolutional Neural Network, a pedestrian in an infrared image can be effectively recognized, and real-time tracking for a pedestrian target in an infrared video can be achieved.

The embodiment of the invention discloses a pedestrian rerecognition method based on a fusion convolution neural network. The method comprises the steps of: constructing a fusion convolution neural network; preprocessing training pedestrian images and obtaining the convolutional activation map by inputting the images into the fusion convolutional neural network; obtaining the whole feature of thetraining pedestrian image by performing the whole pooling, and obtaining the local feature by performing the local horizontal pooling; learning and optimizing the whole feature and local feature respectively, and training the fused convolution neural network; after preprocessing the test pedestrian images, inputting the image to the fusion convolution neural network, and extracting the whole feature and local feature of the test pedestrian image to obtain the final feature; obtaining the pedestrian recognition results by searching for the pedestrian image matching the final feature in the testset as the target image. The invention makes full use of the advantages of the convolution neural network, learns the whole feature and the local feature of the pedestrian image, and finally fuses the two features to represent the pedestrian image, thereby further improving the matching accuracy of the pedestrian recognition.

The invention provides a method for generating and expanding the pedestrian re-identification data based on a generative network. The method includes the steps of generating a new pedestrian video frame sample by utilizing a video prediction network; generating the end-to-end pedestrian background transformation data by means of a deep generative adversarial network; expanding the breadth and therichness of the pedestrian data set by using different data generation methods; and sending the expanded data set to the feature extraction network, extracting the features and evaluating the performance through the Euclidean distance. According to the method, intra-class and inter-class data expansion of pedestrians is taken into consideration as well, more abundant samples can be generated by combining different generative networks, the expanded data set has good diversity and robustness, the problem of performance loss caused by insufficient number of samples and background interference canbe well solved, and the method has general applicability, and the expanded data set can achieve better performance and efficiency in the next step of pedestrian recognition.

The invention discloses a method and device for detecting the posture of a human body. The method comprises the steps of acquiring a figure image, wherein the figure image comprises a pedestrian area and a background area, and the pedestrian area comprises a left shoulder point and a right shoulder point; detecting the position of the pedestrian area in the figure image and the width of the pedestrian area, and detecting the number of key points and the position of each key point in the figure image; if the number of the key points is equal to the preset number, marking the figure image to be an unblocked state; calculating a first distance between the left shoulder point and the right shoulder point; and if the first distance is not less than a turn-back threshold, marking the figure image to be a non-turn-back state. The state of the figure image is judged according to the number of the key points and the distance between the key points. Classification can be performed on a plurality of figure images according to the state of the figure images, or unblocked figure images without a turn-back state are selected to perform the next step of operations such as pedestrian recognition, so that the accuracy of pedestrian recognition can be improved.

The present invention discloses a video pedestrian recognition method based on a convolution neural network. The method comprises a step of reading the video in a video database, intercepting a video frame, and extracting the HOG feature of the video frame, a step of constructing and training the convolution neural network, a step of selecting a plurality of character feature attributes and designing a support vector machine classifier for each character feature attribute and carrying out training, a step of inputting the HOG feature into a trained convolution neural network model, and carrying out sorting classification on each character feature . The method has the advantages that the method of the convolution neural network is employed to reflect a recognition rate well, the HOG feature is extracted, thus the amount of calculation is reduced, the speed is improved, the constructed convolution neural network has a certain depth, at the same time combined with a support vector machine, the classification is carried out for multiple times, and the recognition efficiency and accuracy are improved greatly.

The invention discloses a pedestrian identification method of a camera network based on multi-level depth feature fusion. A new network model is learnt on a pedestrian database by migrating parameters of a pre-training network to the pedestrian database, a plurality of multi-level depth features are extracted using the new network model, and then a Softmax classifier in the last layer of a convolutional neural network is replaced by an SVM classifier to achieve the purpose of making full use of the multi-layer depth features. Furthermore, the multi-level depth features are used for constructing a plurality of groups of SVM classifiers of binary classification and decision values of these binary classifiers are linearly weighted to obtain final classification results. The invention can effectively improve the accuracy of recognizing a pedestrian target by way of multi-level feature fusion in a decision-making layer of the SVM classifier.

The invention discloses an ROS (Robot Operating System) based robot automatic following method. According to the method, data is acquired by adopting a laser radar, preprocessing is performed on the data, the data is clustered by using a hierarchical clustering algorithm, a pedestrian double-leg model is taken as a pedestrian recognition feature, the position between the legs represents the pedestrian position, and a defect that the laser radar is not obvious in feature and low in recognition rate is solved by a method of resampling. The automatic following method is implemented by reasonably utilizing an ROS, message transfer and function implementation between the parts are facilitated, and a navigation framework of the ROS is utilized to enable a robot to have a certain navigation obstacle avoidance ability in the automatic following process.

A method and a device for recognizing a pedestrian and a vehicle supporting the same are provided. The method includes collecting, by a controller, a far-infrared image using a far-infrared imaging device and detecting a pedestrian candidate group from the far-infrared image. In addition, the method includes extracting, by the controller, pedestrian features based on previously normalized pedestrian database (DB) learning and comparing the pedestrian features with the pedestrian DB learning results to determine similarity. The controller is configured to perform pedestrian recognition based on the comparison result.

The invention discloses a lightweight convolutional neural network pedestrian recognition method, and the method comprises the steps: obtaining original pedestrian data, obtaining a pedestrian image with mark information, and constructing a pedestrian data set; designing an input size of a pedestrian model network, clustering the data set, and selecting an appropriate candidate box; preprocessingthe image, and expanding data; constructing a convolutional neural network, and sending the preprocessed image to the network for training to obtain a network model with a pedestrian recognition function; sending the images with the mark information into a network in batches for training; and checking the loss value and the accuracy on the training set, adjusting the learning rate or increasing the number of iterations and then training again if the result is not ideal, testing the trained network on the verification set if the result is ideal, and adjusting the network again according to theverification result. The identification precision can be improved while the real-time performance of the target detection model is ensured, so that the network model can run on a hardware platform with relatively low configuration.

A method for utilizing multiple detectors to conduct pedestrian detection on video images of scene change comprises the steps of (A) respectively obtaining scene background model of each scene of multiple scenes and respectively training a pedestrian detector of each scene; (B) establishing corresponding relation sets of scene background models and pedestrian detectors; (C) obtaining the video images of scene change and dividing the video images into multiple video clips; (D) obtaining clip background model of each video clip in the multiple video clips and using the pedestrian detectors determined scene based on the background models of the video clips to detect pedestrians in each video clip. By means of the method, pedestrian recognition rate during scene change can be effectively improved, and labor cost is reduced.

The invention discloses a safety helmet wearing detection method and device based on single-model prediction. The method comprises the following steps: inputting an original image into a deep convolutional neural network, extracting apparent characteristics of the original image from different layers of the deep convolutional neural network, and acquiring characteristic graphs of different scalesfrom the apparent characteristics by adopting a characteristic pyramid network; respectively inputting the feature maps of different scales into a coordinate regression network and a pedestrian recognition network, respectively outputting the position of a pedestrian target detected in the original image and the confidence of recognition, finding an optimal target bounding box through a non-maximum suppression method, and eliminating redundant bounding boxes; inputting the feature maps of different scales into a safety helmet wearing classification network based on an attention mechanism, andfinally obtaining a detection result whether the pedestrian target wears the safety helmet or not. Whether a worker wears a safety helmet or not in workplaces such as a factory area and a constructionsite shot by a monitoring camera is accurately recognized through a single model.

The invention relates to a pedestrian attribute recognition method and system, computer equipment and a storage medium, and the method comprises the steps: extracting an image frame from a real-time video, inputting the image frame into a trained target detection model, and obtaining a pedestrian image outputted by the trained target detection model; inputting the pedestrian image into a trained pedestrian recognition model, and obtaining a classification result output by the trained pedestrian recognition model according to preset pedestrian attributes, wherein the preset pedestrian attributes comprise a gender attribute and an age attribute; and obtaining the number and proportion of pedestrians with different gender attributes and different age attributes according to the classificationresult. According to the invention, the pedestrian images can be extracted from the real-time video, the data of the pedestrian images are classified and counted to obtain pedestrian flow informationwith different attributes, and the popularity degree and popular groups of the sightseeing scenic spot are indirectly displayed by monitoring the pedestrian flow information with different attributes, and the traffic smoothness and the commercial service range can be reasonably planned.

The invention provides a pedestrian recognition system, a recognition method, a computer readable storage medium. The pedestrian recognition system comprises a database module which stores calibratedface data; a camera/photographing device used for acquiring pedestrian information; a topology module used for acquiring the spatial distribution of the camera/photographic device to form a spatial model; a face recognition module used for comparing the pedestrian information with calibrated face data and screening target pedestrians; a pedestrian re-identification module used for extracting signinformation of a target pedestrian in the pedestrian information and an identification number of the camera/camera device; and a verification module used for acquiring the sign information, the identification number and the spatial model, and identifying and drawing the travel track of the target pedestrian according to the spatial model and the sign information. Through real-time comparison and combination of technologies of face recognition, pedestrian re-recognition and intelligent pedestrian track judgment, the recognition accuracy is ensured, and compared with global monitoring camera searching, the comparison time is greatly shortened, and the false detection condition is reduced.

The invention belongs to the technical field of image processing, and relates to a pedestrian re-recognition method based on multi-channel consistency features. The method includes the following stepsthat N to-be-matched image pairs of training data and test data and labels ln corresponding to the image pairs are input, wherein n equals to 1,...,N; semantic feature representations of the input image data and color texture spatial distribution feature representations are extracted; consistency feature representations of the semantic feature representations and the color texture spatial distribution feature representations are obtained through multi-scale feature matching; a binary classifier is built for the obtained consistency feature representations, and a probabilistic representation for the same target is output. The method has the advantages that pedestrian recognition is carried out through comprehensive pedestrian image semantic attributes and color distribution features, and the method is high in precision, stable in performance and suitable for solving the problem of pedestrian re-recognition in a complex scene.

The invention relates to a pedestrian detection and recognition method based on a deep learning cascade neural network, and the method comprises the steps: (1) sending a preprocessed video image sequence to a first-level neural network, and obtaining the original information of a pedestrian in an image; (2) segmenting a local image of the pedestrian in the image and carrying out normalization processing to construct a pedestrian recognition data set; and (3) sending the pedestrian recognition data set to a second-level neural network, and extracting feature information of the pedestrians to realize identity recognition of the pedestrians. According to the method, the problems of inaccurate target positioning, low pedestrian resolution, low pedestrian identity recognition accuracy and the like in the image are solved, relatively good image information of the target pedestrian can be obtained, and the pedestrian detection and recognition accuracy is improved. The method is good in practice effect and high in operation speed, detection and identity recognition of the target pedestrian can be rapidly and accurately achieved in real time, and the method is suitable for various fields ofvideo monitoring, intelligent communities, specific place supervision and the like.

The invention relates to an intersection pedestrian recognition safety control system and method based on short-range communication. The system comprises an information acquiring unit, an image processing unit, a control unit, a pedestrian detection early-warning unit and a communication unit, wherein the control unit is connected with the information acquiring unit, pedestrian and vehicle images acquired by the information acquiring unit are transmitted to an image recognition unit, and acquired traffic signal information is transmitted to the control unit; the image processing unit is used for acquiring and processing an image of the recognition device and comparing the image with a characteristic image of the device so as to obtain a recognition result, and vehicle speed and pedestrian speed obtained through processing are transmitted to the control unit; the control unit is used for performing logic operation on the information basis of the acquired vehicle speed, the pedestrian speed and the traffic signal lamp time. A vehicle and roadside unit communication control module is used for acquiring the traffic state signal in real time, the state of a signal lamp is judged through the control system, and therefore the pedestrian detection early-warning module can be controlled to give an alarm.

The invention discloses a monitoring video pedestrian recognition and tracking method and device and a storage medium, and the method comprises the steps: constructing an SSD target detection model which comprises a MobileNet cascade neural network and a feature extraction network which are trained in advance; decoding a monitoring video to obtain image data, selecting one frame in the image dataand inputting the frame into the SSD target detection model for face detection and recognition to obtain position box information of each pedestrian in the current frame image; and tracking each pedestrian in real time according to the position box information of each pedestrian in the current frame image and generating a moving track. According to the embodiment of the invention, the SSD networktrained by deep learning is adopted, the feature acquisition network for pedestrian recognition and tracking in the monitoring video is optimized, pedestrian face detection is realized through the MobileNet cascaded neural network, the calculation speed and accuracy are greatly improved, real-time pedestrian detection of a camera can be realized, and the reliability of security monitoring is improved.

The invention provides a pedestrian identification method based on image with FHOG-LBPH feature. By statistical average of fusion HOG features (FHOG) , the optimal features are selected according to the combination with single optimal feature and the separable criterion of bhattacharyya distance, and improved FHOG-LBPH feature are obtained through fusion LBPH characteristics, fundamentally reducing the feature dimension; A classifier is obtained by using a support vector machine (SVM) to train the sample characteristics in order to obtain the classification of the test sample. The experimental results show that the method makes the pedestrian detection accuracy and real-time performance a certain improvement. Effectiveness of the method is validated by the image automatically shot, and the method has certain application value in the real pedestrian recognition.

The invention relates to a pedestrian re-identification method based on the fusion of attribute features and weighted block features, comprising the following steps: constructing an attribute featureextraction sub-network, which integrates the manually extracted features and the features extracted by a depth neural network; using A weighted cross-entropy loss function to train the attribute feature extraction subnetwork; constructing A block-based feature extraction sub-network, which can fuse the depth features of multiple blocks. Training the sub-network based on block feature extraction, setting the weighted fusion layer of local loss function, learning different weights independently, and then endowing each local loss function; training the whole network to extract the pedestrian feature representation which combines the attribute feature and the depth feature based on the block. The invention is reasonable in design, effectively combines attribute features and depth features, optimizes the loss function calculation method, obtains a good pedestrian recognition result, and greatly improves the overall matching accuracy of the system.

The invention discloses a multi-pedestrian cross-camera online tracking system. The system comprises a pedestrian detection module, a human body recognition module, a face recognition module, a targetscreening module, a multi-camera target fusion module and a cross-camera target feature collection module. According to the invention, face recognition and human body recognition technologies are fused, so that the recognition accuracy is improved; meanwhile, the spatial position points of historical tracks are fused under the condition of camera crossing to improve the accuracy of pedestrian recognition. Face recognition and spatial position data are fused in the human body recognition technology, the human body recognition result is corrected in real time, the problems that in the prior art, only a single recognition technology is adopted for tracking, a target is prone to being lost, and recognition errors are prone to occurring are solved, and therefore the accuracy of the multi-pedestrian online tracking technology can be effectively improved.