Abstract: The present disclosure discloses a method for encrypting a visually secure image based on adaptive block compressed sensing and non-negative matrix decomposition. Firstly, the Tetrolet transform is performed on the plain image, then the sparsity degree is optimized on the sparsity matrix and the matrix scrambling is performed, such that the sparsity degree in each block region of the image matrix is equalized. Then according to the image information, the sampling number of the block region is calculated, the measurement matrix is constructed and optimized, and the image is compressed by using the optimized measurement matrix. The compressed image is then scrambled and diffused to complete the encryption process. Finally, the image information is embedded into the carrier image through non-negative matrix decomposition to obtain a visually safe ciphertext image. The decryption process is the inverse of the encryption process.

Abstract: The present disclosure relates to a wide-range perpendicular sensitive magnetic sensor and the method for manufacturing the same, the magnetic sensor includes a substrate, a plurality of magnetic tunnel junctions, a plurality of magnetic flux regulators, a first output port and a second output port.

Abstract: A method for constructing an artificial intelligence (AI) integrated model is provided, including: obtaining a training dataset, an initial graph network model, and a plurality of base models; then iteratively training the initial graph network model by using training data in the training dataset and the plurality of base models, to obtain a graph network model; and then constructing the AI integrated model based on the graph network model and the plurality of base models, where an input of the graph network model is a graph structure consisting of outputs of the plurality of base models. Since the graph network model considers neighboring nodes of each node in the graph structure when processing the graph structure, the graph network model fully considers differences and correlations between the base models when fusing the outputs of the plurality of base models.

Abstract: The present disclosure relates to a wide-range perpendicular sensitive magnetic sensor and the method for manufacturing the same, the magnetic sensor includes a substrate, a plurality of magnetic tunnel junctions, a plurality of magnetic flux regulators, a first output port and a second output port.

Abstract: The present disclosure relates to an AC/DC closed-loop current sensor, including a magnetism gathering iron core, a TMR chip, a signal processing circuit, a signal generator, and a feedback coil. The TMR chip is arranged at an air gap of the magnetism gathering iron core and connected to the signal processing circuit. The signal processing circuit is connected to the signal generator. The feedback coil is wound around the magnetism gathering iron core and connected to the signal generator. The signal processing circuit is configured to select from the induced signal of the TMR chip and make an amplification to obtain a current signal component and send the current signal component to the signal generator. The signal generator is configured to adjust a current output to the feedback coil based on the current signal component, and output a measurement result of the selected current signal component.

Abstract: Example encoding/decoding methods, communication apparatus, and systems are provided. One example encoding method includes encoding first data by using an encoding model, to obtain N pieces of encoded data, where N is an integer greater than 2. De-correlation is performed on the N pieces of encoded data to obtain N pieces of de-correlated encoded data, where a part or all of the N pieces of de-correlated encoded data may be for determining second data, the second data is for restoring third data, and the third data is the first data, or an approximation degree between the third data and the first data is greater than a specified threshold The N pieces of de-correlated encoded data are sent.

Abstract: The present disclosure discloses a method for watermarking depth image based on mixed frequency-domain channel attention, relating to the field of artificial neural networks and digital image watermarking; the method includes: step 1: a watermark information processor generating a watermark information feature map; step 2: an encoder generating a watermarked image from a carrier image and a watermark information feature map; step 3: a noise layer taking the watermarked image as an input, and generating a noise image through simulated differentiable noise; step 4: a decoder down-sampling the noise image to recover watermark information; step 5: a countermeasure discriminator classifying the carrier image and the watermarked image such that the encoder generates a watermarked image with a high quality. The present disclosure combines the end-to-end depth watermark model with frequency-domain channel attention to expand an application range of the depth neural network in the field of image watermark.

Abstract: The present invention provides a robot dynamic obstacle avoidance method based on a multimodal spiking neural network. The present invention realizes a robot obstacle avoidance method in a dynamic environment by fusing laser radar data and processed event camera data and combining with the intrinsic learnable threshold of the spiking neural network for a scenario comprising dynamic obstacles. It solves the difficulty of failure of obstacle avoidance due to the difficulty in perceiving the dynamic obstacles in the obstacle avoidance task of a robot. The present invention helps the robot to fully perceive the static information and the dynamic information of the environment, uses the learnable threshold mechanism of the spiking neural network for efficient reinforcement learning training and decision making, and realizes autonomous navigation and obstacle avoidance in the dynamic environment. An event data enhanced model is combined to better adapt to the dynamic environment for obstacle avoidance.

Abstract: A method for segmenting a camouflaged object image based on distraction mining is disclosed. PFNet successively includes a multi-layer feature extractor, a positioning module, and a focusing module. The multi-layer feature extractor uses a traditional feature extraction network to obtain different levels of contextual features; the positioning module first uses RGB feature information to initially determine the position of the camouflaged object in the image; the focusing module mines the information and removes the distraction information based on the image RGB feature information and preliminary position information, and finally determines the boundary of the camouflaged object step by step. The method of the present invention introduces the concept of distraction information into the problem of segmentation of the camouflaged object and develops a new information exploration and distraction information removal strategy to help the segmentation of the camouflaged object image.

Abstract: The present invention belongs to the technical field of computer vision, and provides a video semantic segmentation method based on active learning, comprising an image semantic segmentation module, a data selection module based on the active learning and a label propagation module. The image semantic segmentation module is responsible for segmenting image results and extracting high-level features required by the data selection module; the data selection module selects a data subset with rich information at an image level, and selects pixel blocks to be labeled at a pixel level; and the label propagation module realizes migration from image to video tasks and completes the segmentation result of a video quickly to obtain weakly-supervised data. The present invention can rapidly generate weakly-supervised data sets, reduce the cost of manufacture of the data and optimize the performance of a semantic segmentation network.

Abstract: The invention belongs to scene segmentation's field in computer vision and is a depth-aware method for mirror segmentation. PDNet successively includes a multi-layer feature extractor, a positioning module, and a delineating module. The multi-layer feature extractor uses a traditional feature extraction network to obtain contextual features; the positioning module combines RGB feature information with depth feature information to initially determine the position of the mirror in the image; the delineating module is based on the image RGB feature information, combined with depth information to adjust and determine the boundary of the mirror. This method is the first method that uses both RGB image and depth image to achieve mirror segmentation in an image. The present invention has also been further tested. For mirrors with a large area in a complex environment, the PDNet segmentation results are still excellent, and the results at the boundary of the mirrors are also satisfactory.

Abstract: The present invention discloses an image encryption method based on an improved class boosting scheme, which comprises the following steps: acquiring parameters of a hyperchaotic system according to plaintext image information; generating weights required by class perceptron networks through the plain text image information; bringing the parameters into the hyperchaotic system to obtain chaotic sequences, and shuffling the chaotic sequences by a shuffling algorithm; pre-processing the chaotic sequences after shuffling to obtain a sequence required by encryption: and bringing a plaintext image and the sequence into an improved class boosting scheme to obtain a ciphertext image, wherein the improved class boosting scheme is realized based on the class perception networks. The method solves the problems that update and prediction functions in an original boosting network are too simple and easy to predict or the like, so as to obtain the ciphertext image with higher information entropy.

Abstract: The present invention provides a data fusion and reconstruction method for fine chemical industry safety production based on a virtual knowledge graph. In view of the characteristics of fine chemical industry safety production data, such as a large amount of structured data, a multi-source heterogeneous database and a strong sequential logic, the present invention innovatively proposes a method of using a virtual knowledge graph to complete the fusion and reconstruction of a traditional database for fine chemical industry. The present invention fuses static structured knowledge in the field of fine chemical industry with a real-time dynamic database for chemical industry safety production in the concept of ontologies for the first time to organize time series data in the form of entities. In addition, the mapping rules of the existing OBDA system are improved based on a data set of the present invention.

Abstract: The present invention belongs to the technical field of knowledge graph, and provides a knowledge graph construction method for an ethylene oxide derivatives production process. According to data types and characteristics, data sources of the ethylene oxide derivatives production process are sorted and divided into three types: structural data, unstructured data and other types of data. An ontology layer and a data layer of a knowledge graph are constructed by combining top-down and bottom-up methods. A data-driven incremental ontology modeling method is proposed to ensure the expandability of the knowledge graph. For structural knowledge extraction, the safety of original data storage is ensured by means of virtual knowledge graph, and a new mapping mechanism is proposed to realize data materialization. For unstructured knowledge extraction, an entity extraction task is realized on the basis of a BERT-BiLSTM-CRF named entity recognition model by integrating a pre-training language model BERT.

Abstract: The present disclosure discloses an image encryption method based on multi-scale compressed sensing and a Markov model. According to the difference in information carried by low-frequency coefficients and high-frequency coefficients of an image, different sampling rates are set for the low-frequency coefficients and the high-frequency coefficients of the image, which can effectively improve the reconstruction quality of a decrypted image. The decrypted image obtained by the present disclosure has higher quality than the decrypted image generated by the existing scheme, and a better visual effect and more complete original image information can be obtained.

Abstract: The present invention discloses an image encryption method based on an improved class boosting scheme, which comprises the following steps: acquiring parameters of a hyperchaotic system according to plaintext image information: generating weights required by class perceptron networks through the plain text image information: bringing the parameters into the hyperchaotic system to obtain chaotic sequences, and shuffling the chaotic sequences by a shuffling algorithm: pre-processing the chaotic sequences after shuffling to obtain a sequence required by encryption: and bringing a plaintext image and the sequence into an improved class boosting scheme to obtain a ciphertext image, wherein the improved class boosting scheme is realized based on the class perception networks. The method solves the problems that update and prediction functions in an original boosting network are too simple and easy to predict or the like, so as to obtain the ciphertext image with higher information entropy.

Abstract: The invention belongs to scene segmentation's field in computer vision and is a depth-aware method for mirror segmentation. PDNet successively includes a multi-layer feature extractor, a positioning module, and a delineating module. The multi-layer feature extractor uses a traditional feature extraction network to obtain contextual features; the positioning module combines RGB feature information with depth feature information to initially determine the position of the mirror in the image; the delineating module is based on the image RGB feature information, combined with depth information to adjust and determine the boundary of the mirror. This method is the first method that uses both RGB image and depth image to achieve mirror segmentation in an image. The present invention has also been further tested. For mirrors with a large area in a complex environment, the PDNet segmentation results are still excellent, and the results at the boundary of the mirrors are also satisfactory.

Abstract: A method for segmenting a camouflaged object image based on distraction mining is disclosed. PFNet successively includes a multi-layer feature extractor, a positioning module, and a focusing module. The multi-layer feature extractor uses a traditional feature extraction network to obtain different levels of contextual features; the positioning module first uses RGB feature information to initially determine the position of the camouflaged object in the image; the focusing module mines the information and removes the distraction information based on the image RGB feature information and preliminary position information, and finally determines the boundary of the camouflaged object step by step. The method of the present invention introduces the concept of distraction information into the problem of segmentation of the camouflaged object and develops a new information exploration and distraction information removal strategy to help the segmentation of the camouflaged object image.

Abstract: The present invention belongs to the technical field of computer vision and provides a data active selection method for robot grasping. The core content of the present invention is a data selection strategy module, which shares the feature extraction layer of backbone main network and integrates the features of three receptive fields with different sizes. While making full use of the feature extraction module, the present invention greatly reduces the amount of parameters that need to be added. During the training process of the main grasp method detection network model, the data selection strategy module can be synchronously trained to form an end-to-end model. The present invention makes use of naturally existing labeled and unlabeled labels, and makes full use of the labeled data and the unlabeled data. When the amount of the labeled data is small, the network can still be more fully trained.

Abstract: The present invention belongs to the technical field of computer vision, and provides a video semantic segmentation method based on active learning, comprising an image semantic segmentation module, a data selection module based on the active learning and a label propagation module. The image semantic segmentation module is responsible for segmenting image results and extracting high-level features required by the data selection module; the data selection module selects a data subset with rich information at an image level, and selects pixel blocks to be labeled at a pixel level; and the label propagation module realizes migration from image to video tasks and completes the segmentation result of a video quickly to obtain weakly-supervised data. The present invention can rapidly generate weakly-supervised data sets, reduce the cost of manufacture of the data and optimize the performance of a semantic segmentation network.