WiMi Hologram Cloud a leading global Hologram Augmented Reality (“AR”) Technology provider, announced that it integrated semantic information and lidar technology into the SLAM algorithm to explore the lidar-based semantic segmentation SLAM algorithm in dynamic environments. By performing semantic segmentation, potential moving objects in the environment can be displayed, which helps the SLAM algorithm to filter out the moving objects in the feature tracking and mapping module, thus obtaining more accurate pose estimation results. Semantic information is considered to be important information for robots to move from perceptual to cognitive intelligence. Semantic SLAM is an important approach to integrate semantic information into the environment representation.
Read More about Interview : AiThority Interview with Keri Olson, VP at IBM IT Automation
The SLAM algorithm for semantic segmentation based on lidar first utilizes spatial attention networks for the semantic segmentation of point clouds. The full convolutional neural network can perform efficient feature extraction and classification of the point cloud data, which enables the segmentation of different objects and backgrounds in the environment. Through semantic segmentation, we can get information about the moving objects, which helps to filter out the moving objects in the subsequent feature tracking and map construction process, thus improving the accuracy of pose estimation. In addition, in order to better handle the SLAM problem in dynamic environments, a priori knowledge is introduced to guide the categorization criteria of environment elements. By fully utilizing the existing environment knowledge, we can more accurately determine which elements in the environment are static and which are dynamic. This introduction of a priori knowledge can effectively improve the accuracy of identifying and tracking dynamic elements. After identifying the dynamic elements, gesture estimation and semantic graph construction can be further realized. Posture estimation refers to the speculation of the robot’s position and orientation in the environment by analyzing sensor data. Semantic map construction, on the other hand, utilizes known environment models and semantic segmentation results to generate maps containing semantic information. The completion of these steps can provide accurate information support for intelligent robots to navigate and make decisions in dynamic environments.
The key technology modules of SLAM algorithm based on lidar and semantic constraints proposed by WiMi include several aspects such as lidar data processing, semantic segmentation, separation of moving and static objects, attitude estimation, and map construction, which cooperate to jointly solve the challenges of SLAM in dynamic environments and improve the accuracy and reliability of environment perception and self-localization.
Lidar data processing: Lidar is an important sensor for acquiring 3D point cloud data of the environment. In SLAM algorithm, the lidar data is pre-processed with operations such as denoising, filtering and feature extraction. These operations aim to extract useful feature information for subsequent attitude estimation and map construction processes.
Recommended: NovelVox Introduces Intuitive Visual IVR for Cisco, Avaya, and Genesys Contact Centers
Separation of moving and static objects: In dynamic environments, there is a need to separate dynamic objects from static backgrounds. This step is usually based on semantic segmentation results and a priori knowledge to distinguish dynamic objects from static backgrounds by comparing the motion properties of the point cloud data with known environment models. This effectively filters out the influence of dynamic objects and improves the accuracy of attitude estimation and map construction.
Position estimation: Posture estimation is used to infer the position and orientation of the robot in the environment by analyzing the sensor data. The lidar-based SLAM algorithm usually uses Extended Kalman Filter (EKF) or nonlinear optimization methods for attitude estimation. Semantic constraint information can be used to constrain the pose estimation process to improve its accuracy and robustness.
Map construction: Map construction is the integration of sensor data acquired in the environment into a 3D map. Lidar-based SLAM algorithm can generate maps containing semantic labels by integrating lidar data with semantic information. Such a map can provide more information about the structure and features of the environment, providing a richer context for navigation and decision-making of intelligent robots.
With the development of AI, 5G and IoT, the problem of perception and localization of intelligent robots in dynamic environments has become a popular research field. WiMi’s lidar-based SLAM algorithm combines the techniques of FCN and semantic segmentation, which can effectively solve the challenges of SLAM in dynamic environments, and has the technical advantages of accuracy, robustness, scene understanding and real-time performance. With the advantages of accuracy, robustness, scene understanding and real-time, it not only has important application prospects in the fields of industrial automation, intelligent transportation and robot navigation, but also provides strong support for the development of intelligent robots in the future, so that the intelligent robots will be able to achieve a higher level of perception, understanding and response capabilities in dynamic environments, and bring more convenience and safety to people’s lives.
Recommended AI News: HUB International Embeds Auto Insurance into Wayne Reaves Auto Dealer Management Software
[To share your insights with us, please write to sghosh@martechseries.com]
Comments are closed.