Abstract: A point cloud data processing method, and a robot and a robot control method using the same are provided. The method includes: obtaining image data including an RGB image and a depth image that is collected through an RGBD camera; obtaining an original mask image by segmenting out targets from the RGB image using a target segmentation mode; obtaining an optimized mask image by performing a pixel-level processing on the mask image; obtaining, based on the optimized mask image and the depth image, a plane equation of each of the targets in the optimized mask image; performing, using the plane equation of each of the targets, a depth value assignment on a plane position of the target in the optimized mask image that is not assigned with the depth value; and obtaining target point cloud data by performing a point cloud conversion on the depth image.

Abstract: A method for area dividing in a map for a mobile robot includes: obtaining a binary map corresponding to a target area; determining one or more first areas in the binary map according to partition information, wherein the one or more first areas are undivided areas in the binary map, the partition information corresponds to second areas in the binary map, and the second areas are divided areas in the binary map; for each first area of the one or more first areas, determining, according to size of the first area, the first area as a newly added second area, or merging the first area into a third area that is one of the second areas, thereby obtaining a target area-dividing map; and controlling the mobile robot according to the target area-dividing map.

Abstract: A method for extracting video features may include: obtaining a target video sequence that comprises a number of video frames; performing video frame feature extraction on the target video sequence to obtain video frame features of each of the video frames; performing feature weight calculation on each of the video frame features to obtain the feature weight of each of the video frame features; wherein the feature weight of each of the video frame features is determined by the video frame features of all of the video frames in the target video sequence; and performing feature weighting on each of the video frame features according to the feature weight of each of the video frame features to obtain video features of the target video sequence.

Abstract: An image quality assessing method, an electronic device, and a computer-readable storage medium are provided. The method includes: obtaining a to-be-assessed original image; obtaining a grayscale image and a histogram equalization image corresponding to the obtained original image by performing an image conversion on the original image; calculating an image similarity between the obtained grayscale image and the obtained histogram equalization image; and determining an image quality assessment result of the original image according to the calculated image similarity.

Abstract: A linkage mechanism includes: a base member; a first link rotatably connected to the base member, the first link defining a first arc-shaped guide groove centered on a pivot axis about which the first link rotates relative to the base member; a second link rotatably connected to the first link; a connecting member rotatably connected to the base member and the second link; an actuating mechanism including a linear actuator and a transmission member that is driven by the linear actuator, the transmission member having a first end rotatably connected to the output shaft, and a second end slidably received in the first arc-shaped guide groove. When the linear actuator drives the connecting member to extend and move, the second end of the transmission member abuts against one end of the first arc-shaped guide groove, which drives the first link to rotate relative to the base member.

Abstract: A method for determining positions of navigation target points includes: acquiring an initial map, wherein the initial map represents a plan view corresponding to an explored area; performing contour extraction processing on the explored area to obtain a first contour map corresponding to the explored area, wherein the first contour map includes one or more connecting sections, and the one or more connecting sections are between adjacent ones of a number of sub-areas in the explored area, respectively; performing dilation processing on the first contour map to obtain a second contour map to remove the connecting sections; based on the initial map and the second contour map, extracting at least one navigation target point area from the explored area; and determining the positions of the navigation target points based on connected components corresponding to the at least one navigation target point area.

Abstract: A method for synthesizing a talking head video includes: obtaining speech data to be synthesized and observation data, wherein the observation data is data obtained through observation other than the speech data; performing feature extraction on the speech data to obtain speech features corresponding to the speech data, and performing feature extraction on the observation data to obtain non-speech features corresponding to the observation data; performing temporal modeling on the speech features and first non-speech features to obtain low-dimensional representations, wherein the first non-speech features are non-speech features that are sensitive to temporal changes; and performing video synthesis based on the low-dimensional representations and second non-speech features, wherein the second non-speech features are non-speech features insensitive to temporal changes.

Abstract: A storage medium, a robot, and a method for generating navigation map are provided. By disposing a first lidar and a second lidar located higher than the first lidar, it constructs a first map corresponding to the first lidar based on first laser data collected by the first lidar, and calculate second positioning data corresponding to the second lidar during constructing the first map, constructs a second map corresponding to the second lidar based on the second positioning data and second laser data collected by the second lidar, and obtains a navigation map corresponding to the robot by fusing the first map with the second map, such that the fused map includes not only positioning information provided by the first map, but also obstacle information provided by the first map and the second map.

Abstract: A method for obtaining a feature extraction model, a method for human fall detection and a terminal device are provided. The method for human fall detection includes: inputting a human body image into a feature extraction model for feature extraction to obtain a target image feature; in response to a distance between the target image feature and a pre-stored mean value of standing category image features being greater than or equal to a preset distance threshold, determining that the human body image is a human falling image; and in response to the distance being less than the preset distance threshold, determining that the human body image is a human standing image. The feature extraction model is obtained based on constraint training to aggregate standing category image features and separate falling category image features from the standing category image features.

Abstract: An obstacle avoidance control method for a robot includes: obtaining, by a depth camera perception system, envelope motion information of each movable obstacle in a target operating environment in a current control cycle, and accessing a pre-stored actual voxel block position of each environmental obstacle in the target operating environment, wherein the actual voxel block position of each environmental obstacle was measured by the depth camera perception system before the movable obstacles were present in the target operating environment; and determining a number of target obstacles that need to be avoided by a number of key obstacle-avoidance parts of a target robot in the target operating environment according to all of the envelope motion information and all of the actual voxel block positions.

Abstract: A target identification method includes: obtaining an image containing a target to be identified; performing feature extraction on the image to obtain image features in the image; and inputting the image features into a target identification network model to obtain an identification result that determines a class to which the target to be identified belongs. The target recognition network model includes a loss function that is to constrain a first distance corresponding to each triplet in a number of triplets and a second distance corresponding to each triplet in the triplets. The first distance represents a distance between anchor image features and positive sample image features in each triplet in the triplets, and the second distance represents a distance between the anchor image features and a class center mean of a number of classes in each triplet in the triplets.

Abstract: A robot control method, and a computer-readable storage medium and a wheel-legged biped robot using the same are provided. The method includes: determining a kinetic model of the wheel-legged biped robot; determining, using the kinetic model, a sliding surface of the wheel-legged biped robot; determining, according to the sliding surface, a double power reaching law and a sliding mode control law of the wheel-legged biped robot; and controlling, according to the sliding surface, the double power reaching law and the sliding mode control law, the wheel-legged biped robot. Through the above-mentioned method, the adaptability of the wheel-legged biped robot to uncertain external disturbances can be enhanced, thereby improving its robustness to effectively maintain its balance even in the environment with complex terrain.

Abstract: A method includes: obtaining heat maps including a predetermined number of key points of a human body; performing depth separable convolution on a feature map corresponding to one of the heat maps corresponding to each of the key points and a convolution kernel of a corresponding channel of the human body posture recognition model to determine a key point feature map corresponding to each channel of the human body posture recognition model; performing local feature fusion processing and/or global feature fusion processing on the key point feature map corresponding to each channel to obtain fusion posture feature maps; determining a linear relationship between the channels of the human body posture recognition model based on the fusion posture feature maps; and updating weight coefficients of the corresponding channels of the human body posture recognition model by using the linear relationship between the channels of the human body posture recognition model.

Abstract: A method for avoiding singularities of a robotic arm includes: calculating a virtual environment external force required by the robotic arm to avoid singularities based on current joint positions of joints on the robotic arm when it is determined that the robotic arm needs to avoid singularities; obtaining a current end force of the robotic arm and a desired end trajectory of the robotic arm; performing admittance control calculation based on the virtual environment external force, the current end force and the desired end trajectory to obtain a corrected end trajectory of the robotic arm; and controlling the robotic arm to move based on the corrected end trajectory.

Abstract: A user exercise detection method applicable in a robot includes: obtaining first measurement data from at least one inertial measurement unit (IMU) sensor that is arranged at a designated body part of the user, and detecting a posture of the user relative to the robot based on the first measurement data; obtaining second measurement data from the at least one IMU sensor, and determining whether an exercise of the user corresponding to the posture is detected according to a preset threshold parameter and the second measurement data; in response to detection of the exercise, obtaining exercise data when the user performs the exercise multiple times through the at least one IMU sensor; and adjusting the threshold parameter according to the exercise data.

Abstract: A method for determining a pose of a robot having a lidar including: obtaining a first pose of the robot in a map coordinate system; determining first positions of laser points corresponding to the lidar in the map coordinate system according to the first pose when the lidar performs laser scanning; determining matching scores between the first positions and grids where the first positions are located according to the first positions and mean values of the grids where the first positions are located, wherein the grids are grids in a probability map corresponding to the map coordinate system; determining a first confidence level for the first pose based on the matching scores; and determining a target pose according to the first confidence level.