144results about How to "Simple network structure" patented technology

The invention relates to the field of pattern recognition, computer vision and deep learning, provides a fine edge detection method based on a deep fusion correction network and a fine edge detectiondevice thereof, and aims at solving the problems that edge positioning in the image is insufficiently accurate and the detected edge is insufficiently fine. The method comprises the steps that step S1, the multi-scale features of an input image are acquired through the forward propagation part network of a convolutional neural network; step S2, the final image feature having the same resolution with that of the input image is acquired by using the method of gradually increasing the feature resolution through the reverse correction part network of the convolutional neural network; and step S3,the feature channel of the final image feature is dimensionally reduced into the single channel, and the edge detection result is generated through the fitting function. The network structure is simpler, and the acquired image feature expression further retains detail features so that the detection effect is better and the edge visualization result is finer.

The invention discloses an IP message sending method, a cluster system and a load equalizer. The method comprises the following steps: receiving an IP message sent by external customer equipment; determining a target MAC address of the IP message according to the IP message; directly sending the IP message to a server corresponding to the target MAC address through a link layer; and receiving a response message of the IP message sent by the server, and sending the response message to the customer equipment. The cluster system with higher throughout and simpler networking can be realized through the embodiment of the invention.

The invention discloses a width transfer learning network and a rolling bearing fault diagnosis method based on the same, and belongs to the technical field of bearing fault diagnosis. The invention provides a novel width transfer learning network and a rolling bearing intelligent diagnosis method based on the same, and aims to solve the problems of scarcity of vibration data with mark informationof a rolling bearing under a variable load, large distribution difference between source domain data and target domain data in the same state, unbalanced distribution of multi-state data and low diagnosis accuracy and model training efficiency. According to the invention, a width learning system is utilized to extract features of source domain data and target domain data and construct a sample set, and on the basis, a balanced distribution adaptation method in transfer learning is adopted to reduce the difference between a source domain and a target domain. A chicken swarm algorithm is introduced to optimize width transfer learning network parameters and establishing a width transfer learning network model. The proposed network model is applied to rolling bearing fault intelligent diagnosis under the variable load, and an experimental result verifies the high efficiency and accuracy of the proposed method.

The invention discloses a method for estimating the number of persons based on deep learning, and belongs to the population density estimation based on deep learning. According to the invention, the method comprises the steps: employing a single-column convolution neural network based on a convolution layer and a pooling layer; learning the population features through the training of a large number of samples, estimating the population density map of an input image, carrying out the integration of the density map, and achieving the estimation of the number of persons on the image. Compared with other conventional deep learning algorithm, the convolution neural network employed in the invention is simple in structure, is low in complexity, is short in training time, and is higher in estimation precision.

The invention discloses a quick evolution method for an optimized deep convolution neural network structure. The method comprises the following steps that: 1) utilizing an evolution algorithm based onGNP (Gene Network Coding) to effectively construct a nonlinear CNN (Convolutional Neural Network) structure, and carrying out mutation on various hyper-parameters of the CNN structure to search an optimal CNN hyper-parameter combination; 2) in an evolution process, designing a multi-objective network structure evaluation method, simultaneously taking classification accuracy and the complexity degree of the classifier as an optimization objective so as to aim at generating the CNN classifier with high classification accuracy and a simple structure; and 3) putting forwarding an incremental training method, and carrying out filial generation CNN structure training on the basis of a previous generation of CNN structure. By use of the method, the training frequency of the model can be reduced,and the time complexity of an algorithm is lowered.

The invention discloses a pedestrian re-recognition method based on global feature splicing. The network performance is improved from the spatial dimension. Firstly, an SE-ResNeXt50 network is used asa backbone network to extract pedestrian image features; splicing the features extracted from different convolution layers to enable the feature information to be complementary; finally, convolutionprocessing is conducted on the spliced features again, and features with high fine granularity are obtained, wherein the introduced clustering loss function is different from a triple loss function commonly used at present, and the clustering loss function and the cross entropy loss function are combined to train a model for the first time; according to the method, the recognition effect is basedon a Market1501 data set, two evaluation indexes of Rank-1 and mAP reach 95.9% and 94.6% respectively, the current recognition effect is high, the network structure is simple, and the calculated amount is small.

The invention discloses an image blind motion blur removing method based on a cyclic multi-scale generative adversarial network. According to the method, a cyclic multi-scale encoder and a decoder areused as generators, and a corresponding decision device is constructed. And the adversarial loss, the multi-scale mean square error and the multi-scale gradient error of the generated image and the clear image are taken as a loss function of the generative adversarial network, and the loss function is optimized by a gradient descent method. According to the method, the generative adversarial network is used to learn the relationship between the motion blurred image and the corresponding clear image, and a complex blurred kernel estimation process is omitted. According to the method, the edgefeatures of the image can be extracted, a simpler network structure and fewer parameters are achieved, the network model is easier to train, and the restoration effect is good.

The invention provides an electric link bus system. In the invention, central air conditioning equipment, electrical switch equipment, lighting equipment, energy management equipment, safety protection equipment, environmental monitoring equipment and the like in a building can be integrated, and all the equipment can be monitored and managed in a centralized way by using a unified bus network structure and a control platform; meanwhile, interconnection and communication with other systems can be realized by communication interfaces, thus, a modern electric link network system is formed.

The invention provides a short message transmitting method of one-set-multiple-number service based on a MAP signaling. The technical proposal adopted by the invention is that: after a one-set-multiple-number platform receives a short message transmitted with an auxiliary number by a submitted one-set-multiple-number user, the short message is simulated into a switching center MSC, the short message is transmitted to an auxiliary number local short message service center through the MAP signaling triggering a MO short message process; the subsequent short message transmission is processed by the auxiliary number local short message service center, and finally transmitted to the mobile telephone of a receiver. The invention does not fulfills the messages transmission through short message gateways, but fulfills the message transmission by submitting the message to the auxiliary number short message service center through the mode of the MAP signaling; other network elements of the prior communication network do not need any modifications, the prior services of the prior operators can not be affected after implementation, and the using habits of the mobile users of the prior network can not be affected either. The invention well improves the short message function of the one-set-multiple-number service.

The invention relates to a virtual private networking system and a data processing method thereof. The method comprises the steps as follows: a first service provider border router receives virtual private network identity sent by a second service provider border router in advance; the first service provider border router receives a first data packet and encapsulates the first data packet with an IP packet header; the virtual private network identity is added into the first data packet, and the data packet is sent to the second service provider border router; the second service provider border router receives the data packet and carries out route looking-up and transmitting according to the virtual private network identity. The system comprises the first service provider border router and the second service provider border router. By adopting the system and the data processing method of the invention, IP technology can be directly utilized to realize the virtual private network, which reduces the requirement for network equipment and operating costs.

The invention relates to a system for realizing first color bell service with soft exchange technique, and relative method, wherein said system at least comprises: application server AS, media server MRS, and web server which are connected by local network to form service platform as application layer; it uses soft exchanger SS as control layer, and uses the IP grouped network with confirmed service quality QoS as carrier layer, and uses the soft exchange network formed by signal gateway SG of local network PSTN and relay gateway TG as access layer, and relative PSTN; the soft exchange network is one increment service platform independent from local network PSTN, and spanning local network, to connect the PSTN via SG and TG, to realize the first color bell service via special number section. The invention has simple structure and lower cost.

The invention discloses a fine-grained image recognition method and system based on block detector and feature fusion convolutional neural network. The method comprises the following steps of: firstly, obtaining local features and global features; directly classifying the global features to obtain loss1; respectively filtering the local features and the global features by using a block detector toobtain local filtering features and global filtering features; carrying out global maximum pooling on the obtained local filtering features to obtain local discriminative feature blocks, wherein thelocal discriminative feature blocks can be directly classified to obtain loss2; in addition, constructing a feature fusion flow to fuse the local filtering features and the global filtering features,obtaining a hierarchical multi-layer representation, and obtaining loss3 through direct classification; introducing an attention cross-layer pooling method to carry out filtering supervision on the network to obtain a loss4; and finally, performing weighted summation on the four loses to obtain total loses, so that fine-grained image recognition can be effectively realized.

The invention provides a method and device for converting media streams. The method comprises the steps as follows: a media server receives the media streams borne by E1 signals; the media server acquires data information in a slot time of each path of E1 signals in the media streams borne by the E1 signals; and the media server encapsulates the data information into real-time transport protocol (RTP) packets, and the media steams borne by the E1 signals are converted into RTP data streams. According to the invention, the unified processing for a public switched telephone network (PSTN) and polyphonic ringtone/video polyphonic ringtone signals of wireless networks can be realized, and the mutual conversion of medial signals in a circuit switching domain and a packet switching domain on the media server can be achieved, thus simplifying a network structure, and eliminating the gateways on media parts; and the conversion results are configured in accordance with the demands of the media server, thus improving the processing efficiency and saving considerable resources.

The invention relates to a blending agent access control method capable of facing wireless industrial network, comprising the following steps: building various kinds of equipment in the wireless industrial network into reticular and star-shaped mixed topology structure; on the basis of the mixed topology structure, defining structuring superframe; according to the above reticular and star-shaped mixed topology structure and the superframe structure, scheduling communication resources according to two stages characterized by being from a gateway to the route equipment and from the route equipment to the site equipment; according to the above defined superframe structure, adopting the strategy combining a fixed channel and frequency hopping to communicate. The invention adopts reticular and star-shaped mixed topology structure to simplify network structure, lower maintenance and management difficulty, improve system flexibility and improve network reliability, has favourable real-time and low energy consumption and improves network compatibility; on the other hand, adaptive channel allocation and frequency hopping improve the reliability of the network and improve the flexibility and the expandability of the network.

The invention relates to an optical passive network protecting system, comprising: a main optical network terminal, a standby optical network terminal, a fiber circuit protecting structure, an optical splitter, an optical network unit, wherein the main optical network terminal is connected with an input end 1 or an input end 2 through the fiber circuit protecting structure. The standby optical network terminal is connected with an input end 1 or an input end 2 of the optical splitter through the fiber circuit protecting structure. The output end of the optical splitter is connected with the optical interface of the optical network unit. Compared with the traditional technique, the invention provides a plurality of protections for the PON system, with lower cost.

The present disclosure provides a real-time vehicle overload detection method based on a convolutional neural network (CNN). The present disclosure detects a road driving vehicle in real time with a CNN method and a you only look once (YOLO)-V3 detection algorithm, detects the number of wheels to obtain the number of axles, detects a relative wheelbase, compares the number of axles and the relative wheelbase with a national vehicle load standard to obtain a maximum load of the vehicle, and compares the maximum load with an actual load measured by a piezoelectric sensor under the vehicle, thereby implementing real-time vehicle overload detection. The present disclosure has desirable real-time detection, can implement no-parking vehicle overload detection on the road, and avoids potential traffic congestions and road traffic accidents.

The invention relates to the technical field of risk control of the Internet financial industry, in particular to a CS-PNN-based customer credit risk assessment method and system. Compared with BP andRBF neural networks, the PNN fuses a Bayesian decision theory and density function estimation on the basis of a radial basis function, the method has the advantages of simple network structure, few adjustment parameters, short operation time, no local minimum point and the like; compared with GA, PSO, ACO and other optimization algorithms, the CS algorithm searches for a global optimal solution by simulating the combination of the parasitic propagation behavior of the valley bird nest and the Levy flight search principle, has the advantages of being few in parameter setting, high in convergence speed, high in universality and robustness, easy to implement and the like, and can efficiently balance local search and global search of the algorithm; the CSPNN model obtained by optimizing the smoothing factor of the PNN by the CS has the advantages of simple network structure, high convergence rate, good fault tolerance, high robustness, high classification accuracy, strong sample appendingcapability and the like, and can meet the requirement of real-time credit risk assessment of a loan system.

The invention discloses a reverse communication method of a crossing safety zone, a device and a system. The method includes sending explore packets to file gateways of isolation devices, obtaining working state information of the isolation devices through echo packets of the explore packets, comparing the working state information of the isolation devices, and choosing one corresponding gateway of one of the isolation devices to transmit a reverse communication packet according to comparison results. By means of the method, the device and the system, a communication bus crossing a safety zone can be set up, and the structure of a network can be simplified. One of working states of the isolation devices is chosen, the communication packets are coordinated for transmission, and the resource utilization rate of equipment is improved. Exchange visits of systems are benefited, and therefore system performance, expansibility and reliability are improved.

The invention discloses a method and device for obtaining a posture classification model. The method comprises the following steps carrying out the posture recognition of an image accommodating a preset posture through a key point detection model, and obtaining a posture label of the image accommodating the preset posture; and performing model training according to the image accommodating the predetermined attitude and the attitude label to obtain an attitude classification model. The attitude classification model obtained by using the method is simpler in network structure and lower in requirement on calculation resources, and can be operated on mobile equipment in real time; the human body posture recognition process of the posture classification model is only a posture recognition matching process, human body skeleton key point detection is not needed, the recognition process is simple, the recognition efficiency is high, and the processing time of a single-frame image is shortened.

The invention relates to the technical field of urban electric power heating, in particular to a computer control system for urban electric power central heating and a control method thereof. The control system comprises a command control center, a region work station and a main control system; the command control center is connected with the region work station through internet, and the region work station is connected with the main control system of a user terminal. The control method comprises the following steps of: (1) starting a program and verifying; (2) reading data from a service center, and meanwhile starting a data server; and (3) waiting for a control command of a user, if the command of the user exists, executing the command of the user, and uploading an execution result to the data server, if no command of the user exists, executing a routing inspection command and uploading a routing inspection result to the data server. The invention can realize the interaction between electric energy and heat energy and central control of the system, is beneficial to peak load shifting of a power grid, has low electricity consumption cost, realizes the utilization of electricity at the valley period while solving the problem of urban heat supply, and achieves the aim of peak shaving and valley filling.