AHY-SLAM: Toward Faster and More Accurate Visual SLAM in Dynamic Scenes Using Homogenized Feature Extraction and Object Detection Method

Han Gong,Lei Gong,Tianbing Ma,Zhicheng Sun,Liang Li

DOI: https://doi.org/10.3390/s23094241

IF: 3.9

2023-04-25

Sensors

Abstract:At present, SLAM is widely used in all kinds of dynamic scenes. It is difficult to distinguish dynamic targets in scenes using traditional visual SLAM. In the matching process, dynamic points are incorrectly added to the pose calculation with the camera, resulting in low precision and poor robustness in the pose estimation. This paper proposes a new dynamic scene visual SLAM algorithm based on adaptive threshold homogenized feature extraction and YOLOv5 object detection, named AHY-SLAM. This new method adds three new modules based on ORB-SLAM2: a keyframe selection module, a threshold calculation module, and an object detection module. The optical flow method is used to screen keyframes for each frame input in AHY-SLAM. An adaptive threshold is used to extract feature points for keyframes, and dynamic points are eliminated with YOLOv5. Compared with ORB-SLAM2, AHY-SLAM has significantly improved pose estimation accuracy over multiple dynamic scene sequences in the TUM open dataset, and the absolute pose estimation accuracy can be increased by up to 97%. Compared with other dynamic scene SLAM algorithms, the speed of AHY-SLAM is also significantly improved under a guarantee of acceptable accuracy.

engineering, electrical & electronic,chemistry, analytical,instruments & instrumentation

What problem does this paper attempt to address?

The paper mainly addresses the issues encountered when applying visual SLAM (Simultaneous Localization And Mapping) technology in dynamic scenes and proposes a new algorithm—AHY-SLAM.

The core problem the paper aims to solve is that traditional visual SLAM methods find it difficult to distinguish dynamic objects in the scene, which can lead to the erroneous inclusion of dynamic points in the camera pose calculation, thereby reducing the accuracy and robustness of pose estimation.

To solve the above problems, the paper proposes the following innovations:

1. **AHY-SLAM Algorithm**: This is a new SLAM algorithm based on adaptive threshold homogenized feature extraction and object detection methods. It adds three new modules to ORB-SLAM2: keyframe selection module, threshold calculation module, and object detection module. These improvements enhance the accuracy and robustness of pose estimation in dynamic scenes.

2. **Keyframe Selection Method**: Introduces a keyframe selection strategy based on the optical flow method. This method uses the LK optical flow method to preprocess the input ordinary frames to determine whether they should be selected as keyframes. Only images selected as keyframes will undergo ORB feature point extraction, thus reducing the computational overhead of extracting feature points for every frame.

3. **Adaptive Threshold Homogenized Feature Extraction Algorithm**: Adjusts the threshold of FAST corner detection according to the grayscale values of different regions of the image, making the distribution of feature points on the image more uniform and avoiding the concentration of feature points on dynamic objects.

Through the above improvements, tests on the TUM public dataset show that AHY-SLAM can reduce the absolute trajectory error by up to 97% in multiple dynamic scene sequences compared to ORB-SLAM2. Additionally, compared to other dynamic scene SLAM algorithms, it significantly improves speed while ensuring acceptable accuracy.