In intelligent mobile robot technology, SLAM (simultaneous localization and map construction) technology is the key component, which enables the robot to achieve autonomous navigation and understanding of the surrounding environment in the unknown environment. Although traditional SLAM methods usually assume that the environment is static or slowly changing, the dynamic nature of the environment, especially the presence of dynamic objects, poses significant challenges to SLAM algorithms in practical application scenarios such as drone navigation, autonomous driving, and robot exploration.To address this challenge, We put forward a sturdy semantic visual SLAM algorithm designated as DFE-SLAM. The innovation of DFE-SLAM is that it combines semantic segmentation network and optical flow pyramid technology to effectively reduce the influence of dynamic targets on positioning accuracy, thus significantly improving the accuracy of positioning in dynamic environments. In addition, we introduced Nerf mapping technology to make our SALM building maps clearer. To verify the performance of DFE-SLAM, we conducted extensive experiments on the TUM RGB-D dataset as well as in real-world environments. The experimental results show that the absolute ballistic accuracy of DFE-SLAM is significantly improved compared with that of ORB-SLAM3. This result not only validates the effectiveness of DFE-SLAM in dynamic environments, but also provides strong support for future research and application.
The paper introduces a factory-specific SLAM algorithm that seamlessly integrates deep learning with feature point filtering to address challenges associated with inaccurate and biased positioning data in industrial environments. By carefully considering the distinct characteristics of factory settings, particularly those where AGV robots operate, our approach effectively distinguishes between dynamic and static entities commonly encountered in such environments. To achieve this, we employ a deep learning-based dynamic object detection mechanism along with a refined feature point filtering process. Initially, deep learning algorithms are utilized to identify potential dynamic objects in the scene, providing valuable prior information. Subsequently, a feature point filtering algorithm is meticulously crafted to eliminate feature points that may introduce interference. This refinement ensures a more rational removal of dynamic feature points, thereby improving the positioning precision and robustness of the visual SLAM system in dynamic factory environments. Extensive experimental results demonstrate that, when compared to ORB-SLAM2 and DS-SLAM, the proposed algorithm achieves superior positioning and mapping accuracy in factory settings. This advancement not only addresses a longstanding challenge in robotics but also represents a significant stride towards enhancing the autonomy and reliability of AGV robots in industrial applications.
Intelligent optical sensing technologies play important roles in many fields, one of which is to help unmanned devices such as UAVs, autonomous mobile robots and intelligent robots to achieve accurate localization and mapping. With the advancement of Industry 4.0 and intelligent manufacturing, the use of autonomous mobile robots has become an important indicator of a country's industrial modernization. As the core issue in the research of autonomous mobile robots technology, autonomous localization and mapping technology has been the focus and difficulty of many scholars at present. Through the efforts of early researchers and engineers, the localization and mapping technology of autonomous mobile robots in simple static environment has achieved fruitful results, and is also playing an important role in the practical industrial application of autonomous mobile robots. However, when the autonomous mobile robots are faced with more complex or changing surrounding environment, the traditional localization and mapping methods based on geometric features such as points and lines can not achieve more accurate results, and even produce many wrong data to hinder the normal operation of the autonomous mobile robots. In this paper, combined with the characteristics of the complex dynamic environment that autonomous mobile robots will encounter in actual work, we propose a method to obtain and utilize the relatively advanced semantic information in the surrounding environment and use it for autonomous mobile robot localization and mapping. The method of this paper uses deep learning technology to mine more advanced semantic information based on the traditional method obtaining the geometric information of the environment, so that the autonomous mobile robot can generate advanced recognition and cognition of the objects inthesurrounding environment, thus assisting it to complete more accurate localization and mapping.
Intelligent optical detecting tracking technologies play important roles in many fields, one of which is to help unmanned devices such as UAVs, autonomous vehicle and intelligent robots to achieve accurate localization and mapping. For medical and nursing robots, the first step in participating in the treatment and nursing process is to accurately locate their location in the ward, and perceive the surrounding environment of the ward. However, when faced with more complex or constantly changing surrounding environments, especially when medical and nursing robots facing a large flow of medical personnel and patients in wards, the hospital environment is relatively complex, then traditional positioning and mapping methods based on geometric features such as points and lines cannot achieve accurate results for medical nursing robots. In this paper, combined with the characteristics of complex dynamic environments encountered in actual wards, we propose a method to obtain high-level semantic information in the surrounding environment and use it for medical and nursing robot’s localization and mapping. Experiments have shown that the semantic based SLAM technology proposed in this article can help medical and nursing robots achieve more accurate localization and mapping results compared to the current popular SLAM technologies, and the use of semantic information can also enable medical and nursing robots to recognize medical devices, laying the foundation for performing other higher-level tasks.
Object tracking is a core subject in computer vision and has significant meaning in both theory and practice. We propose a tracking method in which a robust discriminative classifier is built based on both object and context information. In this method, we consider multiple frames of local invariant features on and around the object and construct the object template and context template. To overcome the limitation of the invariant representations, we also design a nonparametric learning algorithm using transitive matching perspective transformation. This learning algorithm can keep adding object appearance and can avoid improper updating when occlusions appear. We also analyze the asymptotic stability of our method and prove its drift-free capability in long-term tracking. Extensive experiments using challenging publicly available video sequences that cover most of the critical conditions in tracking demonstrate the enhanced strength and robustness of our method.
Object tracking is a core subject in computer vision and has significant meaning in both theory and practice. In this paper, we propose a novel tracking method, in which a robust discriminative classifier is built basing on both object and context information. In this method, we consider multiple frames of local invariant features on and around the object, and construct the object template and context template. To overcome the limitation of the invariant representations, we also design a non-parametric learning algorithm using transitive matching perspective transformation, which is called as LUPT (Learning Using Perspective Transformation). This learning algorithm can keep adding new object appearance into the object template and avoid improper updating when occlusions appear. In this paper, we also analyze the asymptotic stability of our method and prove its drift-free capability in long term tracking. Extensive experiments using challenging publicly available video sequences that cover most of the critical conditions in tracking demonstrate the enhanced strength and robustness of our method. Moreover, in comparison with several state-of -the-art tracking systems, our method shows superior performance in most of cases, especially in long time sequences.
This article [J. Electron. Imaging 25(6), 061602 (2016), doi: 10.1117/1.JEI.25.6.061602] was retracted on 18 December 2018 due to double publication in this and another peer-reviewed journal. The authors regret this mistake.
To overcome the drawback that Boosting decision trees perform fast speed in the test time while the training process is relatively too slow to meet the requirements of applications with real-time learning, we propose a fast decision trees training method by pruning those noneffective features in advance. And basing on this method, we also design a fast Boosting decision trees training algorithm. Firstly, we analyze the structure of each decision trees node, and prove that the classification error of each node has a bound through derivation. Then, by using the error boundary to prune non-effective features in the early stage, we greatly accelerate the decision tree training process, and would not affect the training results at all. Finally, the decision tree accelerated training method is integrated into the general Boosting process forming a fast boosting decision trees training algorithm. This algorithm is not a new variant of Boosting, on the contrary, it should be used in conjunction with existing Boosting algorithms to achieve more training acceleration. To test the algorithm’s speedup performance and performance combined with other accelerated algorithms, the original AdaBoost and two typical acceleration algorithms LazyBoost and StochasticBoost were respectively used in conjunction with this algorithm into three fast versions, and their classification performance was tested by using the Lsis face database which contained 12788 images. Experimental results reveal that this fast algorithm can achieve more than double training speedup without affecting the results of the trained classifier, and can be combined with other acceleration algorithms. Key words: Boosting algorithm, decision trees, classifier training, preliminary classification error, face detection
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.