11 results about "Graphic recognition" patented technology

The application discloses a kind of based on multi-task learning's light hand-drawn figure recognition and parameter regression method and system, including hand-drawn track point sequence is preprocessed into standardization binary image, high-level semantic feature is extracted by feature extraction network, and class and parameter are output in parallel by classification branch and multi-parameter regression branch, call special regression subnetwork refinement regression for pentagon, pentacle, finally output shape type and geometric parameter according to JSON protocol.Model uses two-stage training, mask MSE regression is done to polygon, curve, circular class in joint stage, independent standard MSE regression is done to pentagon, pentacle in special stage, combined with weight cross-entropy classification loss and multi-branch selective back propagation strategy, the application realizes light, efficient hand-drawn figure real-time identification and regular, applicable to mobile terminal image editing class application.

The present invention is a process and structure of a ferroelectric neural network cell with self-learning capability and enhanced character recognition length. The ferroelectric neural network cell is a content addressable memory (CAM) structure (or FeCAM for short) based on a ferroelectric field-effect transistor (FeFET), and the FeCAM is a 1N1P-FeCAM including an N-type FeFET and a P-type FeFET. Therefore, the present invention reduces the number of devices to minimize cell area, and each cell consists of two complementary FeFETs to achieve a density that is 33% higher than that of Ternary CAM (TCAM). H2 plasma treatment (HPT) is effective in enhancing symmetry between ION1 and ION2 to increase the number of mismatch cells (Max HD, MHD). Therefore, the present invention is useful to graphic recognition and can be applied to various fields of artificial intelligence.

The present disclosure relates to a hand-drawn figure recognition method, device and system, and a computer readable storage medium. The hand-drawn figure recognition method comprises: receiving an input of a hand-drawn figure, the input comprising a stroke set, the stroke set comprising at least one stroke, each stroke comprising a sequence of information of a plurality of track points, and the information of each track point comprising coordinates of the track point and writing time; determining whether the stroke set constitutes a closed figure according to coordinate data of the strokes of the input; and in the case where the stroke set constitutes a closed figure, identifying a category of the closed figure based on the coordinate data of the strokes.

The application discloses an intelligent control system and method for glue line image-text recognition and automatic waste disposal, and relates to the technical field of computer vision. The system comprises an image acquisition module, an image preprocessing module, a glue line area positioning module, an image-text recognition module, a defect detection module, an intelligent decision module and a waste disposal execution module. The image acquisition module is used for acquiring image information of a printed matter to be detected. The image preprocessing module is used for carrying out denoising and enhancement processing on the acquired image. The glue line area positioning module is used for identifying and positioning a region where the glue line is located. The image-text recognition module is used for recognizing image-text information in the glue line region. The defect detection module is used for detecting printing defects in the glue line region. The intelligent decision module is used for generating a waste disposal decision according to the recognition result and the defect detection result. The waste disposal execution module is used for executing an automatic waste disposal operation according to the waste disposal decision.

The present disclosure relates to the technical field of artificial intelligence, and particularly relates to a seal identification method and device for business audit, equipment, medium and product, the method comprises: performing character recognition on a seal of a client to be audited to obtain recognition accuracy; performing graphic recognition on the seal to obtain a graphic recognition result; wherein the graphic recognition is used to identify the shape and logo of the seal; based on the recognition accuracy and the graphic recognition result, the comprehensive weights of each recognition model in a model library are calculated; wherein the model library comprises a plurality of recognition models constructed using different algorithms; according to the N recognition models with the largest comprehensive weights, intelligent recognition is performed on the seal; wherein N is greater than or equal to 1. By calculating the comprehensive weights of each recognition model, the most suitable recognition model that matches the seal can be determined, different seals can be identified by different recognition models, which can avoid being cracked or evaded, and is beneficial to improve the recognition accuracy.

The application discloses an AI-P&ID graph recognition data labeling method, relates to the technical field of graph labeling, and is used for solving the problem that the integrity of pipeline topology is destroyed. A plurality of remodeling areas are obtained by remodeling the areas of a P&ID graph to be measured, and local line widths of the areas are generated. Width compression or expansion compensation is performed on the line of the remodeling area according to the local line width. The preliminary center line is obtained by center line prediction on the processed image, and the skeleton node information is extracted. The local main direction and the adjacent interval length of the skeleton node in each remodeling area are detected. The cross direction difference degree is calculated based on the local main direction, and the skeleton node density is generated according to the adjacent interval length. The connection state of the skeleton node is judged by comprehensively considering the cross direction difference degree and the skeleton node density. In the broken state, the gap backfilling is performed, and the pseudo connection point removal or semantic labeling association is selected according to the number of skeleton nodes in the area, so that the skeleton breakage, the pseudo connection point and the node drift are avoided.

The application belongs to the technical field of computer software algorithm processing, and particularly discloses a matching method based on touch point trajectories, which comprises the following steps: firstly, pre-collecting and collecting coordinate data of graphic trajectories to obtain pre-collected coordinate trajectory data and real-time collected coordinate trajectory data; establishing a model library of graphics needing to be matched; dividing a region needing to be recognized into uniform square grids to sequentially establish a comparison model library and a matching model library; eliminating data obviously not matching the matching model from the comparison model library to obtain an eliminated model library; secondly, when a user inputs a graphic on a touch screen and ends the input, grid marks are established to mark trajectory points, and all trajectories in the eliminated model library are sequentially compared to find a graphic with the minimum value as the matched graphic; and the graphic recognition and matching judgment can be realized in the touch screen program, and the matching effect of the current touch trajectory and the trajectory in the model library is improved to a certain extent.

The embodiment of the invention provides a visual sorting method which comprises the following steps: providing a raw material belt, acquiring material information of the raw material belt, and determining optical characteristics of the raw material belt according to the material information; a recognition pattern is arranged on the raw material belt, and the pattern feature of the recognition pattern is determined by the optical characteristic; the raw material belt is subjected to die cutting, so that the raw material belt forms a die cutting part, and the die cutting part is provided with an identification pattern; providing a visual module, and obtaining image data of the die cutting component through the visual module; collecting image data, judging the position of an identification graph in the image data, marking the image data to form a good product data set and a defective product data set, and expanding the defective product data set through a generative adversarial network; and constructing a deep learning model through the non-defective product data set and the defective product data set, and judging whether the die cutting part is qualified or not through the deep learning model. According to the visual sorting method provided by the embodiment of the invention, the visual detection accuracy of the die cutting part can be improved.

This invention discloses a data annotation method for AI-P&ID graphic recognition, belonging to the field of graphic annotation technology, to solve the problem of pipeline topology integrity being destroyed. It obtains multiple reshaped regions by reshaping the P&ID map under test and generates local linewidths for each region. Based on the local linewidths, it performs width compression or expansion compensation on the lines of the reshaped regions. It then predicts the centerline of the processed image to obtain a preliminary centerline and extracts skeleton node information. It detects the local principal direction and adjacent interval length of skeleton nodes in each reshaped region, calculates the cross-direction difference based on the local principal direction, and generates skeleton node density based on adjacent interval lengths. It combines the cross-direction difference and skeleton node density to determine the connection status of skeleton nodes. In the case of a broken state, it performs gap backfilling and selects to remove pseudo-connection points or semantically annotate and associate them based on the number of skeleton nodes in the region, avoiding skeleton breakage, pseudo-connection points, and node drift.

This utility model relates to the field of identification signs and discloses a visual identification sign for molten iron ladle cars, including a mounting base plate, a number identification block and a graphic identification block detachably mounted on the surface of the mounting base plate; the mounting base plate is fixedly mounted on the outer middle of the frame of the molten iron ladle car, and has a number installation area and a verification area arranged sequentially from left to right on it; the number identification block is located in the number installation area, and the number identification block consists of a first code representing the capacity of the ladle car and a second code for numbering the ladle car; the graphic identification block, representing the capacity of the ladle car, is located in the verification area and serves to verify the capacity of the ladle car with the number identification block. This utility model can identify the capacity of the ladle car using the number identification block and the graphic identification block, and can form a mutual verification, ensuring the accuracy of the ladle car capacity. Furthermore, both the number identification block and the graphic identification block protrude from the surface of the mounting base plate, increasing the anti-interference capability of the sign recognition, and also has the advantage of a long service life.