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

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: Embodiments of the present application disclose an image dehazing method and an image acquiring apparatus implementing image dehazing. In the embodiments of the present application, an RGBIR sensor can be adopted to generate a first RGBIR image and a second RGBIR image, such that the generated images are color images and contain infrared information; thus a finally generated image thus has a good dehazing effect; in addition, the image acquiring apparatus adjusts the first exposure time length and the second exposure time length based on the first RGBIR image and the second RGBIR image generated during each exposure, causing the RGB components and the IR component in the image to be balanced, which guarantees the display effect of the finally generated dehazed image.

Abstract: Embodiments of the present application provide an image fusion apparatus and an image fusion method. The image fusion apparatus includes: a light acquisition device, an image processor, and an image sensor having four types of photosensitive channels. The four types of photosensitive channels include red, green and blue RGB channels and an infrared IR channel. The light acquisition device is configured to filter a spectrum component of a first predetermined wavelength range from incident light to obtain target light. The first predetermined wavelength range is a spectrum wavelength range in which a difference between the responsivities of the RGB channels and the responsivity of the IR channel in the image sensor in the infrared band is higher than a first predetermined threshold. The image sensor is configured to convert the target light into an image signal through the RGB channels and the IR channel.

Abstract: An embodiment of the present application provides an image analysis method and device, and the method includes: receiving target information sent by a target image acquisition device; wherein the target information includes at least: a first image acquired by the target image acquisition device, and the identification information of the target image acquisition device; obtaining target area information corresponding to the identification information of the target image acquisition device according to a correspondence between pre-stored identification information of each image acquisition device and respective area information; in the first image, determining a second image corresponding to the target area information, and analyzing the second image.

Abstract: The embodiments of the present disclosure disclose a method and apparatus for reading or writing disk data. The disk includes a primary metadata block storing metadata, a backup metadata block for backing up metadata, and at least one object block group. Each object block group includes a data block storing object data, and an index block storing an index of the object data. The primary metadata block is located at a head of the disk; and the backup metadata block is located at a tail of the disk. The at least one object block group is located between the primary metadata block and the backup metadata block. In reading or writing object data, the index in the index block, the metadata in the primary metadata block and the metadata backed up in the backup metadata block can be separately operated. By applying the embodiments of the present disclosure, the life and security of the disk are increased.

Abstract: Disclosed is an image analysis method and apparatus, upon a program crash on a computing node, the central node may acquire the image analysis task that results in the crash, split the image analysis task into multiple image analysis sub-tasks, transmit sequentially the image analysis sub-tasks to the computing node until a program crash occurs thereon. The central node then acquires an error image analysis sub-task that results in the crash on the computing node, and discard the error image analysis sub-task. The image analysis method can discard only the crash resulting image from an image analysis task, while other images in the image analysis task are retained. As such, fewer images are lost, so that the duration of a frame loss during video playback can be reduced and the reliability of image analysis can be improved. The approach is applicable for image analysis.

Abstract: Disclosed are data storage, data reading, and data cleaning methods and devices, and a cloud storage system. Storage blocks of storage nodes are grouped into resource pools. A method can include: for each resource pool, determining a first number of occupied storage blocks in the resource pool (S401); determining whether the first number meets a preset data cleaning condition (S402); if so, sending a data cleaning instruction to a storage node where each occupied second storage block in the resource pool is located, such that the storage node cleans the data stored on the second storage block in the storage node (S403). That is to say, the management node determines the number of occupied storage blocks of a resource pool. When the number meets a preset data cleaning condition, the data in the storage block is cleaned. Such a process is simple, reducing the workload of management node.

Abstract: Embodiments of the present application provide a method, apparatus, and electronic device for eliminating distortion of a distorted image. Side information components of a distorted image are generated. The distorted image is resulted from image processing on an original image. Side information components represent distortion features of the distorted image with respect to original image. Distorted image color components of the distorted image and the side information components are input into a pre-established convolutional neural network model for convolution filtering to obtain distortion-eliminated image color components. The convolutional neural network model is obtained through training based on a preset training set. The training set comprises original sample images, distorted image color components of multiple distorted images corresponding to each of the original sample images, and side information components of each of the distorted images.

Abstract: The present application discloses a method and system for processing image signals, the method and system sensing visible light bands of optical signals in an area to be imaged, via a first imaging unit, to generate a first image signal, wherein the first image signal includes color information; sensing at least infrared bands of optical signals in the area to be imaged, via a second imaging unit, to generate a second image signal, wherein the second image signal includes illumination information; and synthesizing the first image signal and the second image signal to generate a synthesized image with color information and illumination information. The present application can generate a synthesized image with a variety of advantages, and can meet the two monitoring requirements in low illumination and haze environment which are focused in the field of security.

Abstract: Embodiments of the present application disclose a target detection method and device, and relate to the technical field of video processing.

Abstract: A deep neural network training method and apparatus and a computer device are provided. The deep neural network training method includes: obtaining task attributes of nodes in a current network layer in a tree-like network topology (S101); performing cluster analysis on the nodes in the current network layer based on the task attributes of the nodes in the current network layer and extracting a common part of task attributes of multiple nodes in a same category as a task attribute for a parent node of the multiple nodes (S102); training a network parameter of each parent node based on a task attribute of this parent node (S103); and determining that training of a deep neural network corresponding to the tree-like network topology is completed, after completion of training of all nodes in all network layers (S104). The operation efficiency of deep learning can be improved through this solution.

Abstract: Embodiments of a surveillance method, apparatus, system, electronic device and computer-readable storage medium are provided. In the surveillance method, a non-visible light image and a target image are obtained, wherein the target image is generated from a visible light signal captured during a capture period of the non-visible light image. The method then detects whether an object is present in the non-visible light image. When an object is detected in the non-visible light image, a second location area of the object in a visible light image is determined according to a first location area of the object in the non-visible light image, such that surveillance of the object is implemented based on the visible light image, wherein the visible light image is an image determined based on the target image.

Abstract: Systems and methods are provided for hybrid video encoding. An example method includes: acquiring image information; extracting a background image based at least in part on the image information; detecting whether the background image is stable; and performing encoding switching between a background-based encoding method and a non-background encoding method based at least in part on the detection.

Abstract: An embodiment of the present application provides an image analysis method and device, and the method includes: receiving target information sent by a target image acquisition device; wherein the target information includes at least: a first image acquired by the target image acquisition device, and the identification information of the target image acquisition device; obtaining target area information corresponding to the identification information of the target image acquisition device according to a correspondence between pre-stored identification information of each image acquisition device and respective area information; in the first image, determining a second image corresponding to the target area information, and analyzing the second image.

Abstract: Embodiments of the present application disclose an image dehazing method and an image acquiring apparatus implementing image dehazing. In the embodiments of the present application, an RGBIR sensor can be adopted to generate a first RGBIR image and a second RGBIR image, such that the generated images are color images and contain infrared information; thus a finally generated image thus has a good dehazing effect; in addition, the image acquiring apparatus adjusts the first exposure time length and the second exposure time length based on the first RGBIR image and the second RGBIR image generated during each exposure, causing the RGB components and the IR component in the image to be balanced, which guarantees the display effect of the finally generated dehazed image.

Abstract: Systems and methods are provided for video display control. An example method includes: determining, by one or more processors, whether a coding frame extracted from a video coding stream is an I Frame; if the coding frame extracted from the video coding stream is an I Frame, acquiring, by the one or more processors, a timestamp of the I Frame and a timestamp of an adjacent coding frame after the I Frame; and controlling, by the one or more processors, display of the I Frame based on the timestamp of the I Frame and the timestamp of the adjacent coding frame after the I Frame.

Abstract: Embodiments of the present application provide a navigation method, device and system. The method is applicable to a server and comprises: receiving current position characteristic information sent by a mobile terminal, and determining a position of an area matching the current position characteristic information based on the current position characteristic information and a locally stored database, and taking the determined position as an initial position; receiving target position characteristic information sent by the mobile terminal, determining a position of an area matching the target position characteristic information based on the target position characteristic information and the database, and taking the determined position as a target position; and determining a first route from the initial position to the target position and displaying the first route on an electronic map of the mobile terminal.

Abstract: Embodiments of the present disclosure provide a method and apparatus for detecting a license plate.

Abstract: The present application discloses a method and a device for image stitching. The method includes: obtaining a first image acquired by a first camera and a second image acquired by a second camera, wherein the first image is an image with visible light information, the second image is an image including depth information and visible light information, the first camera and the second camera are disposed adjacent to each other, and there is an overlapping area between the acquired first image and second image; mapping pixels in the second image to the overlapping area and/or an extended area, wherein the extended area is an image area formed by pixels in the second image which are mapped to the outside of the first image; stitching the mapped pixels in the second image to the overlapping area and/or the extended area based on depth information and visible light information of the pixels in the second image to obtain a stitched image.