Abstract: Embodiments of the present application provide a method and apparatus for determining a trajectory of a human target. The method includes: extracting a target feature of a to-be-tracked human target in a to-be-processed image as a to-be-searched target feature; searching for an acquisition attribute corresponding to the to-be-searched target feature based on a pre-established correspondence between target features and acquisition attributes; determining a trajectory of the to-be-tracked human target based on the found acquisition attribute. It can be seen that in this solution, a face image is not required for determining a trajectory of a person. Even if the acquired face image is not clear, the accuracy of determining the trajectory of the person will not be reduced. Therefore, by the solution of the present application, the accuracy of determining the trajectory of the person is improved.

Abstract: A method for acquiring queuing information includes: acquiring a location and a height of each pixel point in a first video image in a world coordinate system, wherein the first video image is a video image captured by a video camera at a queue area; determining a behavior state of each human body at the area based on the location and the height of each pixel point; and acquiring queuing information based on the behavior state of each human body at the area at a current time.

Abstract: A transparent cover of an optical camera system comprises a curved sidewall portion having an outer contour formed by the rotation of a preset continuous curve. The preset continuous curve comprises a first curve segment and a second curve segment, each having a curvature no more than a first preset value, wherein the first back focus offset of a curved surface corresponding to the first curve segment does not exceed the second preset value, and a second back focus offset of a curved surface corresponding to the second curve segment does not exceed a third preset value. A first line segment is formed by connecting an end point of the preset continuous curve with a center of a circle circumscribing edges of a camera, and an included angle between the first line segment and a boundary line between the first and second curve segments is greater than 0.

Abstract: The embodiments of the present application provide an image segmentation method, an image segmentation apparatus, and a fully convolutional network system. The method includes: acquiring a target image to be processed; acquiring image feature data of the target image; inputting the image feature data into a pre-trained target network far image segmentation to obtain an output; wherein the target network is a fully convolutional network comprising a hybrid context network structure, and the hybrid context network structure is configured to extract a plurality of reference features at a predetermined scale and fuse them into a target feature that matches a scale of a target object in a segmented image; and wherein the target network is trained with sample images containing target objects at different scales; and obtaining an image segmentation result for the target image based on the output.

Abstract: The present application provides an image capture method and an image capture device. The image capture method is applied to the image capture device, which comprises a first chip, zoom lenses, image sensors, zoom motors, and focus motors. The number of the first chip is one; and the number of the zoom lenses are two, comprising a first short zoom lens and a second long zoom lens, wherein the two zoom lenses have different zoom ranges, and there is an overlap range between the zoom ranges of the two zoom lenses. The number of the image sensors is two, and the two image sensors correspond to the two zoom lenses respectively, and the two image sensors are communicatively connected to the first chip respectively. The number of the zoom motors is two, the number of the focus motors is two, and the two zoom motors and the two focus motors are communicatively connected to the first chip respectively.

Abstract: Disclosed are an electronic device, an expansion card assembly and a server device. The electronic device comprises a hard disk slot and the expansion card assembly. The expansion card assembly is arranged in the hard disk slot in a pluggable manner; the hard disk slot comprises a first type of interface and a second type of interface. The first type of interface is used for communicating with the hard disk, and the second type of interface is used for communicating with the expansion card assembly. Therefore, the expansion card assembly is expanded by means of the hard disk slot in the electronic device, so that the expansion requirement of the expansion card assembly is met.

Abstract: A method and apparatus for acquiring people flow information, and a method and apparatus for processing people flow information are provided.

Abstract: The embodiment of the application provides an image processing method and apparatus for target recognition, the method comprises: inputting N frames of images into a quality evaluation network model; determining, with the quality evaluation network model, a feature vector of each of the N frames of images according to an attention weight of a preset sub-region image and M quality evaluation parameters; determining quality evaluation values of the N frames of images according to the feature vectors of the N frames of images; determining a target image or a target vector for target recognition according to the quality evaluation values of the N frames of images. According to the technical scheme provided by the embodiment of the present application, the feature vector of each frame of image is determined based on the attention weight of the sub-region image and M quality evaluation parameters.

Abstract: An object recognition method and apparatus for a deformed image are provided. The method includes: inputting an image into a preset localization network to obtain a plurality of localization parameters for the image, wherein the preset localization network comprises a preset number of convolutional layers, and wherein the plurality of localization parameters are obtained by regressing image features in a feature map that is generated from a convolution operation on the image; performing a spatial transformation on the image based on the plurality of localization parameters to obtain a corrected image; and inputting the corrected image into a preset recognition network to obtain an object classification result for the image. In the process of the neural network based object recognition, the embodiment of the present application first transforms the deformed image that has deformation, and then performs the object recognition on the transformed image.

Abstract: The embodiments of the present application provide a method and apparatus for recognizing an identity of a human target. The method includes: obtaining a to-be-recognized image; extracting a target feature of a to-be-recognized human target in the to-be-recognized image as a to-be-searched target feature; searching for a facial information item corresponding to the to-be-searched target feature based on a pre-established correspondence between target features and facial information items; and determining an identity of the to-be-recognized human target based on the found facial information item. It can be seen that, in this solution, there is no need to extract the facial features in the image, and even if the facial region is not clear or is shaded by other objects in the image, the accuracy of identity recognition is reduced. Therefore, accuracy of identity recognition is improved.

Abstract: Embodiments of the present application provide a license plate location method and apparatus. The method includes: acquiring a to-be-located license plate image; sending the to-be-located license plate image to a target network for license plate location; wherein, the target network comprises a feature extraction layer and a regression layer; extracting, by the feature extraction layer, a feature value of the to-be-located license plate image, and sending the feature value to the regression layer; acquiring a combination of a license plate frame area and a license plate character area, which is obtained by the regression layer based on the feature value; and determining a license plate area of the to-be-located license plate image based on a location area corresponding to the acquired combination.

Abstract: Method for transmitting data, camera, and electronic device. The method includes: obtaining an to-be-sent image and to-be-sent target data; determining a first position for the to-be-sent image in an effective image area of an image frame, and determining a second position for the target data in an effective image area of an image frame; taking the to-be-sent image and the target data as data of the same image frame, and sending the to-be-sent image according to the first position and the target data according to the second position in a data sending mode for the effective image area. Target data is data other than to-be-sent image and coaxial data, and coaxial data is notification information exchanged between a data receiving end and a data sending end. Target data and to-be-sent image can be sent to an electronic device through the same image frame without additional wiring. Facility costs are reduced.

Abstract: An operation method and apparatus for a network layer in a Deep Neural Network are provided.

Abstract: A multi-lens camera and a lens adjustment device, the multi-lens camera includes a main housing assembly, lenses, and a lens adjustment device; the main housing assembly includes a main housing and an inner baffle arranged at the top opening of the main housing; the inner baffle and the main housing constitute a main board cavity; a main board is arranged in the main board cavity; the area of the main board is less than that of the main board cavity, the lenses are mounted on the lens adjustment device, and the overall area of the lenses and the lens adjustment device is within the area of the support region of the inner baffle; the lens adjustment device includes a cylindrical component and at least two lens brackets, the lenses are mounted on the lens brackets, which are connected with the component to drive the lenses to rotate around the component.

Abstract: A method and an apparatus for capturing a target object and a video monitoring device are disclosed. The method includes: detecting target objects in a current panoramic video frame acquired by a panoramic camera; determining detail camera position information corresponding to each of the target objects, and determining a magnification corresponding to the target objects; performing block processing on the target objects to obtain at least one target block; for each target block, identifying first target objects, at edge positions, among target objects included in the target block, and determining, based on the first target objects, the detail camera position information and the magnification corresponding to the target block; for each target block, controlling the detail camera to adjust its position and magnification based on the detail camera position information and the magnification corresponding to the target block, and controlling the adjusted detail camera to capture the target block.

Abstract: A method and apparatus for constructing a 3D scene model are provided. The method includes: acquiring a first point cloud corresponding to a current target scene (S101); determining a first partial point cloud corresponding to a dynamic object from the first point cloud (S102); constructing a foreground 3D model based on the first partial point cloud (S103); and superimposing the foreground 3D model with a background 3D model to obtain a current 3D scene model (S104). As can be seen, it is not necessary to fuse each pixel acquired by each depth camera. Instead, the dynamic object is determined in the point cloud acquired by the depth camera. Only the foreground 3D model corresponding to the dynamic object is superimposed with the background 3D model corresponding to the static objects obtained in advance. This reduces the amount of computation to build a 3D scene model.

Abstract: An extended storage device based on a PCIe bus is provided. The extended storage device includes: a video processing device and a RAID card. The video processing device is connected to the RAID card via a PCIe bus; and the RAID card is constructed by an embedded system on chip (SoC). The video processing device may be connected to the RAID card via the PCIe bus, and virtual hard disk nodes corresponding to respective physical hard disk nodes of the RAID card may be built at the video processing device, so that the video processing device may manage physical hard disk nodes of the RAID card as managing a local hard disk, and then may organize these virtual hard disk nodes into a virtual RAID node, while the complex RAID computing is actually achieved through the SOC of the RAID card without occupying resources of CPU of the video processing device, so that a normal operation of business of the video processing device can be ensured.

Abstract: A variable magnification method, apparatus and a computer-readable storage medium is disclosed. The method comprises: acquiring angle parameters for the current position of a camera after receiving a variable magnification instruction; wherein, the variable magnification instruction is configured for instructing the camera to perform the variable magnification based on a target magnification ratio parameter; querying target historical focusing information based on the angle parameters for the current position, wherein, the target historical focusing information comprises a historical magnification ratio parameter and a historical in-focus parameter of the camera; querying a target variable magnification curve based on the historical magnification ratio parameter and the historical in-focus parameter; and controlling the camera to perform the variable magnification based on the target variable magnification curve.