RGB-D Odometry for Autonomous Lawn Mowing

Ochman, Marcin; Skoczeń, Magda; Krata, Damian; Panek, Marcin; Spyra, Krystian; Pawłowski, A.

doi:10.1007/978-3-030-87897-9_8

Cited by 2 publications

(2 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In recent years, with continuous advancements in computer vision, deep learning, and artificial intelligence technologies, the widespread application of depth sensors [1] and Light Detection and Ranging (LiDAR) [2] has propelled the rapid development of three-dimensional (3D) semantic maps [3]. Robots are playing an increasingly important role in daily life, undertaking various tasks to make our lives more convenient.…”

Section: Introductionmentioning

confidence: 99%

Construction of Three-Dimensional Semantic Maps of Unstructured Lawn Scenes Based on Deep Learning

Xie,

Yan,

Zhang

et al. 2024

Applied Sciences

View full text Add to dashboard Cite

Traditional automatic gardening pruning robots generally employ electronic fences for the delineation of working boundaries. In order to quickly determine the working area of a robot, we combined an improved DeepLabv3+ semantic segmentation model with a simultaneous localization and mapping (SLAM) system to construct a three-dimensional (3D) semantic map. To reduce the computational cost of its future deployment in resource-constrained mobile robots, we replaced the backbone network of DeepLabv3+, ResNet50, with MobileNetV2 to decrease the number of network parameters and improve recognition speed. In addition, we introduced an efficient channel attention network attention mechanism to enhance the accuracy of the neural network, forming an improved Multiclass MobileNetV2 ECA DeepLabv3+ (MM-ED) network model. Through the integration of this model with the SLAM system, the entire framework was able to generate a 3D semantic point cloud map of a lawn working area and convert it into octree and occupancy grid maps, providing technical support for future autonomous robot operation and navigation. We created a lawn dataset containing 7500 images, using our own annotated images as ground truth. This dataset was employed for experimental purposes. Experimental results showed that the proposed MM-ED network model achieved 91.07% and 94.71% for MIoU and MPA metrics, respectively. Using a GTX 3060 Laptop GPU, the frames per second rate reached 27.69, demonstrating superior recognition performance compared to similar semantic segmentation architectures and better adaptation to SLAM systems.

show abstract

Section: Introductionmentioning

confidence: 99%

Construction of Three-Dimensional Semantic Maps of Unstructured Lawn Scenes Based on Deep Learning

Xie,

Yan,

Zhang

et al. 2024

Applied Sciences

View full text Add to dashboard Cite

show abstract

“…D EPTH camera technology has been commercially available for more than a decade and its use in various fields such as manufacturing [1], mobile robotics [2], object detection [3], medicine [4], and scene understanding [5] has become an important part of many applications. 3D geometry acquisition within these fields usually refers to measurement ranges of a few meters with typical required accuracies of a few millimeters up to several centimeters.…”

Section: Introductionmentioning

confidence: 99%

Assessment and Improvement of Distance Measurement Accuracy for Time-of-Flight Cameras

Frangež¹,

Salido-Monzú²

2022

IEEE Trans. Instrum. Meas.

View full text Add to dashboard Cite

Time-of-flight depth cameras are interesting sensors for contact-less 3D metrology because they combine mechanical robustness with independence of ambient lighting conditions. Their actual performance depends on many factors and is hard to predict from data sheets. In this study we investigate the deviations of the distance measurements of a high-end phasebased depth camera. We focus on the impact of (i) self-warming and external temperature, (ii) on range noise as a function of distance and acquisition time, and (iii) on distance-dependent biases. We present the dedicated experimental setups comprising a climate chamber, a calibration bench with a reference interferometer, and a laser tracker that provide controlled conditions and ground truth data. These setups allow investigating the absolute accuracy and mitigating repeatable distance biases by adapting the measurement model based on experimental data. For demonstration, we apply the investigation to two stateof-the-art industrial depth cameras of the same brand and type (Helios Lucid), showing significantly different response to external temperature but similar distance-dependent biases. We adapt the measurement model of one of the cameras for distancedependent inter-pixel biases and demonstrate that the resulting parameters reduce also the distance biases of the other camera by about 80 % to less than 1 mm at ranges of up to 1 m. This indicates the potential for batch error compensation. The paper contributes to better understanding distance deviations of depth cameras and to improving the accuracy of such cameras.

show abstract

RGB-D Odometry for Autonomous Lawn Mowing

Cited by 2 publications

References 16 publications

Construction of Three-Dimensional Semantic Maps of Unstructured Lawn Scenes Based on Deep Learning

Construction of Three-Dimensional Semantic Maps of Unstructured Lawn Scenes Based on Deep Learning

Assessment and Improvement of Distance Measurement Accuracy for Time-of-Flight Cameras

Contact Info

Product

Resources

About