Outdoor Flight Testing of a Pole Inspection UAV Incorporating High-speed Vision

Sa, Inkyu; Hrabar, Stefan; Corke, Peter

doi:10.1007/978-3-319-07488-7_8

Cited by 38 publications

(21 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A representative procedure of incremental model pruning. There exists four iterative steps: (1) evaluating importance of each component in a pre-trained deep model; (2) removing the components that are less important to model inference; (3) finetuning pruned model to compensate potentially temporary degradation in performance; (4) evaluating the fine-tuned model to determine whether pruned model is suitable for deployment. An incremental pruning strategy is preferred to prevent over-pruning.…”

Section: Model Pruningmentioning

confidence: 99%

SlimYOLOv3: Narrower, Faster and Better for Real-Time UAV Applications

Zhang¹,

Zhong²,

Li³

2019

2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW)

210

120

View full text Add to dashboard Cite

Drones or general Unmanned Aerial Vehicles (UAVs), endowed with computer vision function by onboard cameras and embedded systems, have become popular in a wide range of applications. However, real-time scene parsing through object detection running on a UAV platform is very challenging, due to limited memory and computing power of embedded devices. To deal with these challenges, in this paper we propose to learn efficient deep object detectors through channel pruning of convolutional layers. To this end, we enforce channel-level sparsity of convolutional layers by imposing L1 regularization on channel scaling factors and prune less informative feature channels to obtain "slim" object detectors. Based on such approach, we present SlimYOLOv3 with fewer trainable parameters and floating point operations (FLOPs) in comparison of original YOLOv3 (Joseph Redmon et al., 2018) as a promising solution for real-time object detection on UAVs. We evaluate SlimYOLOv3 on VisDrone2018-Det benchmark dataset; compelling results are achieved by SlimYOLOv3 in comparison of unpruned counterpart, including ~90.8% decrease of FLOPs, ~92.0% decline of parameter size, running ~2 times faster and comparable detection accuracy as YOLOv3. Experimental results with different pruning ratios consistently verify that proposed SlimYOLOv3 with narrower structure are more efficient, faster and better than YOLOv3, and thus are more suitable for real-time object detection on UAVs. Our codes are made publicly available at https://github.com/PengyiZhang/SlimYOLOv3. Figure 1. Billion floating point operations (BFLOPs) versus accuracy (mAP) on VisDrone2018-Det benchmark dataset. Enabled by channel pruning, our SlimYOLOv3-SPP3 achieves comparable detection accuracy as YOLOv3 but only requires the equivalent floating point operations as YOLOv3-tiny. Such performance is very competitive in drone applications. Details are given in §5.

show abstract

Section: Model Pruningmentioning

confidence: 99%

SlimYOLOv3: Narrower, Faster and Better for Real-Time UAV Applications

Zhang¹,

Zhong²,

Li³

2019

2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW)

210

120

View full text Add to dashboard Cite

show abstract

“…In this paper, we make use of our previous developed robust line feature tracker [ 18 ] as a front-end vision system, and it is summarized in Section 2.2 . A significant difference to our previous works [ 19 , 20 , 21 ] in which different flying platforms had been utilized is the integration of both PBVS and IBVS systems on the same platform. By doing so, we are able to compare both systems quantitatively.…”

Section: Introductionmentioning

confidence: 85%

Inspection of Pole-Like Structures Using a Visual-Inertial Aided VTOL Platform with Shared Autonomy

Hrabar

Corke

2015

Sensors

Self Cite

View full text Add to dashboard Cite

This paper presents an algorithm and a system for vertical infrastructure inspection using a vertical take-off and landing (VTOL) unmanned aerial vehicle and shared autonomy. Inspecting vertical structures such as light and power distribution poles is a difficult task that is time-consuming, dangerous and expensive. Recently, micro VTOL platforms (i.e., quad-, hexa- and octa-rotors) have been rapidly gaining interest in research, military and even public domains. The unmanned, low-cost and VTOL properties of these platforms make them ideal for situations where inspection would otherwise be time-consuming and/or hazardous to humans. There are, however, challenges involved with developing such an inspection system, for example flying in close proximity to a target while maintaining a fixed stand-off distance from it, being immune to wind gusts and exchanging useful information with the remote user. To overcome these challenges, we require accurate and high-update rate state estimation and high performance controllers to be implemented onboard the vehicle. Ease of control and a live video feed are required for the human operator. We demonstrate a VTOL platform that can operate at close-quarters, whilst maintaining a safe stand-off distance and rejecting environmental disturbances. Two approaches are presented: Position-Based Visual Servoing (PBVS) using an Extended Kalman Filter (EKF) and estimator-free Image-Based Visual Servoing (IBVS). Both use monocular visual, inertia, and sonar data, allowing the approaches to be applied for indoor or GPS-impaired environments. We extensively compare the performances of PBVS and IBVS in terms of accuracy, robustness and computational costs. Results from simulations and indoor/outdoor (day and night) flight experiments demonstrate the system is able to successfully inspect and circumnavigate a vertical pole.

show abstract

“…In order to ensure the stability of the flight control system in the complex airborne vibration environment, especially the stability of the attitude loop control, the multi-rotor UAV puts forward higher requirements on the angular rate measurement range and accuracy of the Microelectromechanical System (MEMS) gyroscope equipped in the navigation control system. A sufficiently large angular rate measurement range is the basic premise for multi-rotor UAV control reliability and anti-jamming capability [5]. The angular rate accuracy and reliability of industrial-grade MEMS gyroscopes [6,7,8] are significantly higher than those of the consumer grade, but its range is generally narrower, which makes it difficult to meet the wide-range measurement requirements of multi-rotor UAV [9].…”

Section: Introductionmentioning

confidence: 99%

Performance Enhancement Method for Angular Rate Measurement Based on Redundant MEMS IMUs

Qian

et al. 2019

Micromachines

View full text Add to dashboard Cite

Aiming at the low-cost, wide-range, and accurate measurement requirement for Microelectromechanical System (MEMS) Inertial Measurement Unit (IMU) on a multi-rotor Unmanned Aerial Vehicle (UAV), the paper designs a heterogeneous parallel redundancy configuration scheme. In redundant MEMS IMUs, a high-cost and small-range MEMS gyroscope is combined with low-cost and large-range MEMS gyroscopes. Then, an adaptive data fusion method of redundant MEMS gyroscopes is proposed. By the designed experiments based on the simulation data and the sensor measurement data, the proposed method has been proved that it can effectively improve the angular rate measurement performance of the multi-rotor UAV and broaden the angular rate measurement range on the basis of saving the configuration cost and volume of the micro IMU.

show abstract

Outdoor Flight Testing of a Pole Inspection UAV Incorporating High-speed Vision

Cited by 38 publications

References 19 publications

SlimYOLOv3: Narrower, Faster and Better for Real-Time UAV Applications

SlimYOLOv3: Narrower, Faster and Better for Real-Time UAV Applications

Inspection of Pole-Like Structures Using a Visual-Inertial Aided VTOL Platform with Shared Autonomy

Performance Enhancement Method for Angular Rate Measurement Based on Redundant MEMS IMUs

Contact Info

Product

Resources

About