152 results about "Traffic sign detection" patented technology

A method for detecting and recognizing at least one traffic sign is disclosed. A video sequence having a plurality of image frames is received. One or more filters are used to measure features in at least one image frame indicative of an object of interest. The measured features are combined and aggregated into a score indicating possible presence of an object. The scores are fused over multiple image frames for a robust detection. If a score indicates possible presence of an object in an area of the image frame, the area is aligned with a model. A determination is then made as to whether the area indicates a traffic sign. If the area indicates a traffic sign, the area is classified into a particular type of traffic sign. The present invention is also directed to training a system to detect and recognize traffic signs.

The invention discloses a detection method of middle, small and dense traffic signs in an automatic driving scene. The method comprises following steps of (1) acquiring video data shot by a vehicle dashcam, extracting pictures from the vehicle dashcam, marking traffic signs in the pictures to form a traffic sign data set consisting of <images, target frames> pairs; (2) data preprocessing: preprocessing the traffic sign data set; (3) using the shallow layer network VGG16 as a main network of R-FCN object detection frameworks; (4) improving the VGG16 network model, by use of the characteristics of the shallow layer, reducing the declining times of a characteristic graph and training the RPN network to extract candidate frames; and (5) improving the VGG16 network model, carrying out characteristic combination on the same group of the characteristics of the shallow layer, inputting the combined characteristics into a following R-FCN detection framework, carrying out classification and frame regression on the candidate frames and finally detecting all the traffic signals in the pictures. According to the invention, a detection problem of traffic signs in the automatic driving scene is solved.

The invention provides a small object detection method based on a sensing generation adversarial network. The method includes the following steps: generating an adversarial network, a condition generation network architecture and a verification network architecture. The process is composed of two sub-networks, namely a generator network and a sensing verification network. The generator network converts the expression of a small object to the super-resolution expression of an original object which is similar with a big object. Residual expression between the big object and the small object is generated through residual learning. The verification network inputs the super-resolution expression, and transmits the input to an adversarial branch and a sensing branch. The generator network is excited to generate super-resolution expression which has high detection precision. According to the invention, the method herein, by reducing the expression difference between small object and big object, improves small object detection, provides more comprehensive monitoring, is conductive to detection, and realizes successful detection of traffic signs and pedestrians.

The invention provides a traffic sign detection method in a natural scene. The method comprises the steps that a detection image shot in the natural scene is acquired; luminance information of the detection image is subjected to statistical analysis, different luminance areas are divided according to grade luminance threshold values, pixel ratios of different luminance areas are calculated respectively, and the image is divided into a dark scene, a bright scene, a backlighting scene and a normal scene according to all the pixel ratios and scene classification threshold values; gamma parameter values are selected according to scene classification results, and an adaptive Gamma enhancement algorithm is adopted to perform image enhancement processing on a classification image; a partitioning algorithm is selected to perform image color partitioning according to different scenes in an RGB color space to obtain suspected target areas; a grayscale image obtained after color partitioning is subjected to binarization processing to obtain the suspected target areas after binarization; and the suspected target areas are screened through a feature screener to position traffic sign regions. Through the method, robustness and real-time performance of sign detection can be both considered.

The invention relates to a traffic sign detection and identification method based on pruning and knowledge distillation. The method comprises the following steps: preparing a data set and carrying outdata enhancement; establishing a network and training the network: establishing a YOLOV3-SPP network, loading parameters of a pre-training model trained in the data set ImageNet, and inputting the cut training set images subjected to data enhancement into the network in batches for forward propagation to obtain a model which is an original YOLOV3-SPP network; sparse training: using a scaling coefficient of a BN layer as a parameter for measuring channel importance, adding an L1 regularization item on the basis of an original target function, after adding the L1 regularization item, performingtraining again until loss convergence, and naming the process as sparse training; pruning according to the threshold value; and obtaining a final model by knowledge distillation.

The invention discloses a traffic sign detection and recognition method based on a convolutional neural network, which belongs to the field of digital image processing and machine learning. The method comprises steps: firstly, an RGB image after pre-processing is converted to HSV color space, and a region of interest is obtained through threshold setting; and then, a two-classification convolutional neural network for distinguishing a traffic sign and a non-traffic sign is designed to judge whether the region of interest is a traffic sign. After the position of a traffic sign is obtained, the traffic sign recognition method based on the convolutional neural network is used, parameters such as the layer number and the characteristic pattern number of the convolutional neural network are adjusted, parameters in the network are learnt through a large amount of training samples, and classes of traffic signs at different positions are further recognized. An experiment shows that the method has good adaptability to deformation, partial occlusion and tilt and the like of the traffic sign, and good performance is presented in aspects of recognition effects and recognition efficiency.

The invention relates to a traffic sign detection method based on a convolutional neural network. The method comprises the following steps: step 1, constructing a traffic sign detection network with classification and positioning separated based on the convolutional neural network; 2, in the training stage, training the constructed traffic sign detection network by adopting an enhanced iterative training method to obtain a traffic sign detection model; and step 3, in the use stage, carrying out target detection on the input image by adopting a separation and fusion prediction method to obtaina traffic sign detection result. According to the method, rapid and accurate traffic sign detection is realized in a complex traffic monitoring scene, the robustness to the environment is high, and the detection accuracy for small-size traffic signs is relatively high.

The invention discloses a residual SSD model-based traffic sign detection and recognition method. The method includes the following steps that: first step, multi-scale segmentation is performed on animage; second step, a residual SSD model is constructed with a residual network ResNet101 adopted as the basic network of the SSD; third step, network training is carried out; and fourth step, detection and identification with generalization capacity are completed. The invention aims to improve the accuracy of the detection of small targets by an existing SSD network, and realize the effective detection and recognition of a plurality of types of signs of different sizes in the real traffic scenes in China.

The invention discloses a multi-characteristic synergic traffic sign detection and identification method which is performed according to the following steps. A color probability model is established for traffic signs with different colors through the images of traffic sign samples and representative colors are determined out of the traffic signs with different colors so as to obtain probability check lists for the representative colors, train and obtain shape classifying devices for traffic signs belonging to different categories and identifying models. For traffic images to be detected, each probability check list for the representative colors is used first to get the probability images of the traffic images, which are then converted to grey scale maps. An MSER algorithm is used to detect the areas in the grey scale maps which change stably and the areas are regarded as potential windows to be picked up that meet the pre-set height-width ratio. The shape classifying devices then determines whether the potential windows to be picked up contain traffic signs or not, and if they do, the identifying models will identify these corresponding shapes. The method can achieve a better detection and identification effect by combining the characteristics of colors and shapes of traffic signs.

The invention discloses a traffic sign detection method in automatic driving based on a YOLOv3 network, and belongs to the field of traffic sign detection. The method solves the problems that an existing YOLOv3 network target detection algorithm is not high in detection precision and the detection speed cannot meet the real-time requirement. According to the invention, an improved loss function isprovided, so that the influence of a large target error on a small target detection effect is reduced, and the detection accuracy of a small-size target is improved. An improved activation function is provided, a negative value is reserved, meanwhile, changes and information propagated to the next layer are reduced, and the robustness of the algorithm to noise is enhanced. The real frames in thetraffic sign data set are clustered by using a K-means algorithm to realize the pre-fetching of a target frame position and accelerate convergence of the network. The detection precision mAP of the traffic sign detection model on a test set reaches 92.88%, the detection speed reaches 35FPS, and the requirement for real-time performance is completely met. The method can be applied to the field of traffic sign detection.

The invention discloses a traffic sign detection and identification method based on collaborative bionic vision in a complex city scene, which includes the following steps: A, obtaining multiple to-be-detected images in a continuous scene; B, obtaining a cluster collaboration map of the to-be-detected image set; C, obtaining an attention saliency map of each to-be-detected image; D, obtaining a collaboration saliency map corresponding to each to-be-detected image; E, locating a sign ROI (Region of Interest); F, carrying out two-level biologically inspired transformation on the sign ROI by using a forward channel; and H, using a feature transformation map and traffic sign template images pre-stored in a database to carry out Pearson correlation calculation to complete the identification ofthe to-be-detected images. In the method, visual processing of a target by the human brain is simulated, and bottom-up visual processing and top-down visual processing processes are integrated. The collaborative nature of global images is considered, so that image location is accurate, and the robustness of identification is strong.

The invention relates to a remote traffic sign detection and recognition method based on F-RCNN, which comprises: 1, preprocessing the traffic sign image sample set; 2, pre-training VGG-16 in F-RCNN;3, inputting the traffic sign training data set to the VGG-16 to complete feature extraction; 4, constructing a fusion feature map; enabling the region generation network RPN in the F-RCNN to performregion generation according to the fusion feature map to obtain a candidate region of the traffic sign; 6, inputting all candidate regions to the RoI-Pooling layer in the F-RCNN to generate a fixed-size feature vector; 7, sending the feature vector to the extreme learning machine network to output the category and location of the traffic sign; 8, training the F-RCNN model by the contribution adaptive loss function; 9, completing the traffic sign detection and identification of the actual scene. The invention realizes the detection and identification of the long-distance traffic sign, and the recognition precision is high.

Disclosed are a method and a device for detecting traffic signs in an input image camera. The method comprises a color space converting step of converting the input image into a HSV color space image; a filtering step of filtering, based on a predetermined pass range of a standard color of each of the traffic signs, the HSV color space image to obtain a filtered image, and then generating one or more connected domains based on one or more regions in the filtered image; a removing step of removing, based on a standard rule of the corresponding traffic sign, at least one of the generated connected domains, not being the corresponding traffic sign, and letting others of the generated connected domains be candidate traffic sign domains; and a recognition step of recognizing, based on a feature of each of the candidate traffic sign domains, the corresponding traffic sign.

The invention relates to a long-distance traffic sign detection and identification method suitable for a vehicle-mounted system. The method comprises the following steps: 1, preprocessing a traffic sign image sample set; 2, constructing a lightweight convolutional neural network, and completing the convolutional feature extraction of the traffic sign; 3, through a channel-spatial attention moduleembedded into the lightweight convolutional neural network, constructing an attention characteristic graph; 4, using a region generation network RPN to generate a candidate region of the target; 5, introducing context region information into the target candidate region generated by the RPN, and enhancing the mark classification characteristics; 6, sending the feature vector into a full connectionlayer, and outputting the category and position of the traffic sign; 7, establishing an attention loss function, and training an FL-CNN model; 8, repeating the steps 2-7 to complete sample training ofthe FL-CNN model; and 9, repeating 2-6 to finish traffic sign detection and identification of the actual scene. According to the invention, long-distance traffic sign detection and identification arerealized, and the precision reaches 92%.

The invention discloses a method for detecting a round traffic sign. The method comprises the steps that S1, sample point division is conducted on color samples according to illumination components to acquire sample subsets, and corresponding color classification templates of the sample subsets are generated after division is conducted; S2, an original image is cut through the color classification templates; S3, two-level shape matching operation is conducted on the shape of each single-communication region in the cut image and the shape of the round traffic sign. According to the method, the spatial distance method that the color classification templates are generated by the color sample points is used for image cutting; the round traffic sign is fast positioned regarding to a non-shielding two-level shape matching system and a shielding two-level shape matching system; the calculated amount of color classification is transferred to a color template generation stage before detection from a detection stage, and color classification precision can be improved through increase of the templates. According to the method, detection can be finished through the partially shielded two-level round shape matching system.

The invention relates to a method for detecting rapid robustness traffic signs on an outdoor bad illumination condition. The method comprises the steps that color histograms of different kinds of traffic signs are established; a probability graph based on the various histograms is generated; traffic sign candidate zones based on the MSER are extracted; non-traffic-sign zones are eliminated. According to the method, the different color histograms on different illumination conditions are established, back projection is conducted to generate the probability graph based on the various histograms of input images, the traffic sign images on different illumination conditions are converted into a unified condition, consistency processing is conducted on MSER feature zones, robustness of an algorithm in bad illumination change is improved, and meanwhile, the detection speed is high. Experiments show that on the weak illumination condition and on the strong illumination condition, the detection accuracy of an existing algorithm is obviously reduced, but the detection accuracy rate of the method keeps over 90%. According to the method for detecting the rapid robustness traffic signs on the outdoor bad illumination condition, the red traffic signs, the yellow traffic signs and the blue traffic signs can be extracted and traffic signs in a white background can also be extracted.

The invention relates to a road traffic sign automatic detection and identification method. The method is realized based on an automatic detection and identification system. The automatic detection and identification system comprises a traffic sign acquisition module, a traffic sign detection and processing module and a traffic sign identification module. The traffic sign acquisition module is used for collecting driving videos; the traffic sign detection and processing module is used for carrying out processing on the collected videos, detecting image frames containing traffic signs and carrying out extraction on regions of interest; and the traffic sign identification module is used for carrying out classification and recognition on images of the extracted regions of interest. The road traffic sign automatic detection and identification method can identify various types of traffic signs, is high in precision, high in identification accuracy and good in robustness, reduces influence of illumination, geometric deformation and rotation and the like, and has better anti-interference capability.

Disclosed herein are a speed limit sign recognition system and method using a front camera which can track a traffic sign continuously appearing in images which are obtained by a photographing speed limit sign located in front of a driver using a camera mounted on the front of the vehicle, recognize an internal numeral of the traffic sign, and inform the driver of an actual limit speed through the recognized numeral, including: an image acquisition unit for acquiring a front image using a front camera; a traffic sign detector for detecting a traffic sign from the acquired front image; a recognition unit for recognizing an internal numeral in the detected traffic sign; a tracking unit for tracking a traffic sign continuously appearing in the front image, and eliminating a temporarily misrecognized object; and a decision unit for determining a result of speed limit recognition on the recognized traffic sign.

The present invention provides a road traffic sign detection and identification method. The method comprises the steps of: extracting a region of interest; performing multi-scale sliding traversal; merging image features; and identifying a traffic sign. The method specifically comprises: performing extraction of a region of interest of the traffic sign on an input to-be-detected image; convertingthe to-be-detected image into a gray-scale image; constructing a binary mask image of the region of interest of the traffic sign; performing multi-scale sliding traversal on the gray-scale image and the binary mask image to obtain position coordinates of a detected target; merging the extracted local texture features, local image region features and global features of the to-be-detected image; andclassifying and identifying the merged image features by using a classifier. The present invention relates to an electronic device and a readable storage medium for performing the road traffic sign detection and identification method; and the present invention also relates to a road traffic sign detection and identification system. The technical scheme of the present invention has a high detection rate, a high identification rate, a fast calculation speed, a low false detection rate and good robustness.

An ECU as a driving assistance device has a traffic sign detection section, a lane entry detection section and a judgment section. The traffic sign detection section detects traffic signs including a speed limit symbol based on front-image data captured by an in-vehicle camera. The lane entry detection section detects that the own vehicle has entered a ramp lane to leave a highway. The judgment section detects an extent of a current driving lane as the ramp lane when following conditions (c1) and (c2) are satisfied after the lane entry detection section detects the entry of the own vehicle into the ramp lane: (c1) not less than two traffic signs are arranged along the current driving lane in forward direction of the vehicle; and (c2) an arrangement of speed limit symbols included in the detected traffic signs satisfy a predetermined speed reduction pattern.

The invention discloses a traffic signal detection and identification method based on a cascaded network, and relates to the technical field of image processing and sign identification. The method comprises the following steps that S1) an original image is collected; S2) the original image is detected to obtain a candidate area which may include a traffic signal or lamp, the trained three-order cascaded network composed of a 12-net, 24-net and 48-net is used for detection in a detection process; and S3) the candidate area is identified, the candidate area without the traffic sign or lamp is eliminated, and concrete types of traffic signs and lamps are obtained. The traffic signs and lamps can be detected at the same time, and the problem, in a present traffic sign detection and identification technology, that missing detection is caused by influence of illumination on a color space, small difference between the traffic signal and the background environment and connection of multiple traffic signs can be solved.

The invention discloses an incremental learning based method for detecting and identifying a traffic sign in a traveling video, and is used for solving the technical problem of poor robustness in an existing traffic sign detection and identification method. The technical scheme is as follows: polymerization channel characteristics are adopted for training an Adaboost classifier, and a detector is improved; then a detection result of the detector serves as an observation value of a Kalman filter to perform motion model based tracing; in the tracing process, a new incremental SVM detector is trained on line; when the detection of an original Adaboost detector is failed due to apparent change of the sign, the online incremental detector is used for carrying out detection, the detection result serves as the observation value of the Kalman filter to be input, and targets incapable of being continuously detected are filtered; and finally, a tracing result of the same physical traffic sign is subjected to weighted voting of scale based Gaussian weight, a final identification result is obtained, and the detection and identification robustness are improved.

A traffic sign detection section in an ECU detects a ramp sign in front image data captured by an in-vehicle camera. The ramp sign includes an exit symbol and a speed limit symbol. The exit symbol provides an indication of a place where the own vehicle leaves a current driving lane and enters a ramp lane connected to the current driving lane. A distance detection section detects a width distance between the own vehicle and the ramp sign. A driver's intention detection section detects driver's intention to change a current driving lane to a ramp lane. A lane entry detection section detects that the own vehicle enters the ramp lane when the distance detected by the distance detection section is smaller than a predetermined distance, a direction to change the current driving lane detected by the driver's intention detection section is equal to a direction designated by the exit symbol in the ramp sign.

The invention discloses a real-time traffic sign detection method based on multi-scale pixel feature fusion, and belongs to the field of deep learning and target detection. Firstly, an image containing a traffic sign is obtained and preprocessed; secondly, inputting the preprocessed image into a MobileNetv2 network to carry out feature extraction; inputting the extracted multi-scale feature map into a pixel feature fusion module for pixel rearrangement, and splicing to generate a fusion feature map with semantic information and detail information; performing down-sampling on the fusion featuremap to obtain six scale feature maps, inputting the six scale feature maps into an efficient channel attention module, and allocating weights to feature channels according to importance degrees; inputting the weighted six-scale feature map into an SSD detection layer to predict the position of the bounding box and the category of the object; and finally, carrying out non-maximum suppression to obtain an optimal traffic sign detection result. According to the method, the real-time performance and the accuracy can be considered when the traffic sign image is detected, and the robustness is veryhigh.

The invention discloses a traffic sign detection algorithm based on an improved YOLO v3 algorithm. According to the method, a feature extraction network capable of maintaining high-resolution representation is designed to replace DarkNet-53 in an original YOLO v3 algorithm, so that the detection precision of a small-size target traffic sign is improved, and the parameter quantity of the algorithmis reduced; attention of the detection algorithm to small and medium-sized targets is increased by fusing the feature maps participating in prediction; and optimizing the loss function by using a GIOUalgorithm and a local loss algorithm. According to the method, the detection accuracy of the small-size traffic sign is improved, and the traffic sign can be rapidly and accurately detected and identified on a complex traffic road.

The invention discloses a deep-learning-based high-precision traffic sign detection method and system. On the basis of combination of the deep learning technology and the high-precision traffic sign detection technology, training is carried out on an SSD network and a convolution neural network; a traffic sign characteristic after overlapped cutting based on a proportion in a video stream is extracted by using the trained SSD network; according to the traffic sign characteristic extracted by the SSD network, the characteristic of the traffic sign characteristic is extracted by using the trained convolution neural network; and the extracted characteristic of the traffic sign characteristic matches characteristics of positive and negative type traffic signs in a traffic sign image detection database and the positive type traffic signs are kept, thereby obtaining a high-precision traffic sign matching screening result. Therefore, the accuracy rate of the high-precision traffic sign detection can be improved effectively.

The invention relates to a traffic sign detection identification method, apparatus and system. A sign plate region and a non sign plate region in a to-be-identified traffic scene image are segmented by comprehensive use of an SVF color space and an HSV color space, so that the influence of an interference color can be effectively eliminated; a color channel image is subjected to shape detection and locating, so that the detection of sign plates in multiple shapes such as a round shape, a triangular shape, a rectangular shape and the like can be realized, sign plate locating is realized, and identification types are added; and finally based on an image feature extraction algorithm, sign plate image features are extracted and transmitted to a trained preset classifier for performing sign plate classification judgment. The traffic sign plates erected above lanes or on two sides of a road can be detected and identified in real time in a running process of an intelligent vehicle; then an identification result is sent to a decision-making system of the intelligent vehicle; and the intelligent vehicle pre-judges a traffic condition in front of the vehicle in advance and makes response actions of reducing the speed, turning on a vehicle lamp and the like, thereby ensuring unblocked road and running safety of the intelligent vehicle.

The invention, which belongs to the technical field of machine learning and deep learning, discloses a traffic sign detection and recognition method based on a self-built neural network. According tothe method, on the basis of a shot image, a non-real area of interest of a traffic sign in the image is obtained by using color segmentation in a digital image theory; a real area of interest is obtained by using an SVM classifier; and then the real area of interest is inputted into a self-built convolutional neural network for identification and classification. Therefore, the usage state of the traffic sign can be identified and classified quickly and accurately; the real-time requirement of quick, reliable, and accurate identification is realized; and the time and effort are saved.

The invention discloses a traffic sign detecting method based on a self-adaptation threshold value. Firstly, a reddening and bluing method is adopted by the traffic sign detecting method to carry out image preprocessing, a red-blue grayscale image is obtained, and then multiple times of thresholding processing are carried out on the grayscale image. During each time of thresholding processing, contour detection is carried out on a binary image generated by the thresholding processing; and then, according to shape characteristics of a traffic sign, sign contour matching with geometric condition constraint and Hu invariant moment is carried out, and after filtering and screening are carried out, a suspected traffic sign regional set in a current threshold image is obtained; finally, results of the multiple times of thresholding processing are merged, and according to the frequency for detecting of contour regions of the sign, interest regions of the traffic sign are finally confirmed. According to the traffic sign detecting method, a better thresholding processing effect is provided for candidate areas of the traffic sign in the image, the feature of threshold value self-adaptation is provided, the efficient detecting efficiency and the time performance are provided, and the problem of adaptability under different illuminating conditions is excellently solved.