4results about How to "Avoid tracking" patented technology

The invention provides a communication method and system for a terminal and a cloud server, and the method comprises the steps: the terminal configures an agent client, sets an agent parameter at the agent client, starts an agent, and enters a first communication mode, and the agent parameter comprises a local agent port; in the first communication mode, the terminal applies for a service key from a cloud server, and if the service key is successfully acquired, the terminal is allowed to enter a cloud computer interface; wherein the service key is used for encrypting or decrypting a cloud resource service flow packet transmitted by the local agent port. A terminal proxy client is set to start a proxy, a secure communication mode between a terminal and a cloud server is entered, and a cloud resource service flow packet is encrypted by using service keys held at two ends, so that cloud resources are prevented from being intercepted, leaked and tampered in transmission, and office security is protected; encryption at a flow layer is realized through an agent, the execution efficiency is high, encryption and decryption are implemented by combining a symmetric cryptographic algorithm with a high encryption and decryption processing rate, so that the processing capability and the response speed of the whole system are extremely high, and non-inductive secure acquisition of cloud resources is realized; encryption is carried out on the flow layer, and compared with a service layer, keys can be saved; the terminal is set to obtain the service key and enter the cloud computer interface, so that unauthorized users or hostile attackers are prevented from operating the cloud computer, and the system security is improved.

This invention provides a deep learning-based method for multi-target pedestrian tracking, relating to the field of computer vision technology. The method includes: S1 acquiring video data, extracting frames from the video data to convert them into image frames, and preprocessing the image frames; S2 using an improved YOLOv5-s model to detect pedestrian targets in the image frames, obtaining the pedestrian targets; S3 using an improved Strong SORT model to track the pedestrian targets in each frame, assigning an ID to each tracked pedestrian, and generating pedestrian tracking results. This invention can accurately locate each pedestrian in a video and track their trajectory.

The application discloses a kind of multi-interface adaptive switching control method and system suitable for vehicle charging, after vehicle gun is inserted, the electrical characteristic vector of all physical interfaces is synchronously collected and input one-dimensional convolution neural network, the activation probability of each interface is calculated, if the maximum probability exceeds threshold value, corresponding protocol stack is preloaded and polling is skipped;If confidence is insufficient, a multi-armed bandit model based on Thompson sampling is used to dynamically select the optimal protocol for handshaking;After successful handshaking, the vehicle digital identity is extracted through VIN hash value or physical layer timing fingerprint, the cross-interface shared cache is queried, if hit, the historical charging strategy is directly inherited, otherwise, complete negotiation is performed and a file is created, and the current is output according to the established or negotiated strategy during charging.The application overcomes the inherent defects of existing multi-interface charging pile protocol selection, such as low efficiency and inability to share cross-interface information, and has made significant progress in charging start speed, adaptive ability, user experience and the like.

The invention relates to a two-phase channel turbulence simulation method, which comprises the following steps that: an initial channel turbulence field is established, and the initial channel turbulence field comprises a first direction initial speed, a second direction initial speed and a third direction initial speed; under the condition that at least one liquid drop is added at a specific position, a liquid drop channel turbulent flow field of the at least one liquid drop is initialized; solving the phase field model by using a discrete unified gas kinematics format (DUGKS), and processing a convection term by using a weighted essential oscillation-free format (WENO) to describe two-phase evolution of the turbulent flow field of the liquid drop channel; the first direction target speed, the second direction target speed and the third direction target speed of any volume unit and the sequence parameters of the two-phase flow are obtained through calculation. According to the method, the phase field model is solved by using the DUGKS method, and the WENO format is combined, so that the interface physical and microscopic physical effects can be better captured, and the accuracy and reliability of two-phase flow numerical simulation are improved.