Lars Kuhnert

Structure overview of vegetation detection. A novel approach for efficient vegetation detection using an active lighting system

Fully autonomous navigation has been widely investigated for several decade of years; however, a safe and reliable navigation is still a daunting challenge in terrains containing vegetation. To improve the mobility capability of recent autonomous navigation systems, an additional vegetation detection function has been proposed. Since many proposals of generating vegetation classifier as well as suggestions of using different sensors to implement the function exist, a structured overview is required for vegetation detection in the context of outdoor navigation. Therefore, this paper studies and compares the accuracy and efficiency of existing vegetation detection approaches in a structured way. Furthermore, a new vision system set-up which combines CMOS sensor and Photo Mixer Device sensor with a near-infrared lighting system is also introduced to simultaneously provide depth, near-infrared and color images at high frame rate. Those near-infrared and color information are then used to compute vegetation index or train vegetation classifier to completely realize a real-time robust vegetation detection system. In this paper, a modification of the normalized difference vegetation index is devised, which is then defined as the new standard form of vegetation index for such vision system integrated with an additional lighting system. Finally, we will show the out-performance of the proposed approach in comparison with more conventional ones.

A novel approach for a double-check of passable vegetation detection in autonomous ground vehicles

The paper introduces an active way to detect vegetation which is at front of the vehicle in order to give a better decision-making in navigation. Blowing devices are to be used for creating strong wind to effect vegetation. Motion compensation and motion detection techniques are applied to detect foreground objects which are presumably judged as vegetation. The approach enables a double-check process for vegetation detection which was done by a multi-spectral approach, but more emphasizing on the purpose of passable vegetation detection. In all real world experiments we carried out, our approach yields a detection accuracy of over 98%. We furthermore illustrate how the active way can improve the autonomous navigation capabilities of autonomous ground vehicles.

Terrain classification based on structure for autonomous navigation in complex environments

One of the main challenges for autonomous navigation in cluttered outdoor environments is to determine which obstacles can be driven over and which need to be avoided. Especially in off-road driving, the aim is not only to recognize the lethal obstacles on the vehicles way at all costs, but also to predict the scene category thereby giving a better decision-making framework for vehicle navigation. This paper studies terrain classification based on structure relying on sparse 3-D data from LADAR mobility sensors. While most of recent methods for LADAR processing are purely found on the local point density and spatial distribution of the 3-D point cloud directly. We, on the other hand, introduce a new approach to analyze the point cloud by considering local properties and distance variation of pixels inside edgeless areas. First of all, the edgeless areas are extracted from segmenting the 3-D point cloud into homogeneous regions by Graph-Cut technique. Secondly, the neighbor distance variation inside edgeless areas (NDVIE) features are obtained by calculating the euclidean distance of neighbor distance variation inside each region. Through extensive experiments, we demonstrate that this feature has properties complementary to the conditional local point statistics features traditionally used for point cloud analysis, and show significant improvement in classification performance for tasks relevant to outdoor navigation.

UAV Based Laser Measurement for Vegetation Control at High-Voltage Transmission Lines

Trees which are growing beneath high-voltage transmission lines have to be cut down if the distance to the conductor rope gets too close. It is very hard to find those if only single trees are affected. Chopping simply all trees under a transmission line, which was done in former times, is not a desirable solution for environmental and economic reasons. In this paper a solution is shown which uses an unmanned aerial vehicle and a laser scanner to generate a 3d point-cloud representing the trees and the conductor ropes. Using this point-cloud the decision for cutting down a specific tree can easily be made.

Multiple templates auto exposure control based on luminance histogram for onboard camera

In order to adjust the exposure of high resolution and frequency images from an on-board camera in real time, a simple exposure control approach based on luminance histogram analysis is presented to realize light control precisely and rapidly. The algorithm divides the initial image into nine blocks and calculates the mean brightness of every part. It uses the luminance histogram to analyze the degree of exposure of both: the global image and every part, especially interest areas. According to the area size of the neighborhood on the two edges of the histogram, the different weight matrix can be decided to quickly calculate the optimal brightness for the next frame, and then the optimal exposure time can be determined accurately. The algorithm does not need to determine an expensive quadratic function to decide the weight of every part so as to fast perform auto exposure control. The experimental results show that the algorithm can correctly and fast adjust the exposure time of the on-board camera on an autonomous robot, and has strong adaptability for various shootings outdoor.

Absolute High-Precision Localisation of an Unmanned Ground Vehicle by Using Real-Time Aerial Video Imagery for Geo-referenced Orthophoto Registration

This paper describes an absolute localisation method for an unmanned ground vehicle (UGV) if GPS is unavailable for the vehicle. The basic idea is to combine an unmanned aerial vehicle (UAV) to the ground vehicle and use it as an external sensor platform to achieve an absolute localisation of the robotic team. Beside the discussion of the rather naive method directly using the GPS position of the aerial robot to deduce the ground robot’s position the main focus of this paper lies on the indirect usage of the telemetry data of the aerial robot combined with live video images of an onboard camera to realise a registration of local video images with apriori registered orthophotos. This yields to a precise driftless absolute localisation of the unmanned ground vehicle. Experiments with our robotic team (AMOR and PSYCHE) successfully verify this approach.

A novel approach of terrain classification for outdoor automobile navigation

One of the main challenges for autonomous navigation is to determine which obstacles can be driven over and which need to be avoided. The investigation of reconstructing 3D model of the viewed scene has showed good performance in environments such as in yard, hall way or on road. However, in cluttered outdoor environments where frequently the scenes are unknown and the objects are no more static and rigid, the only use of 3D-point analysis is not sufficient to give good decision for safe navigation. Therefore, we on the other hand address a new approach which reconstructs completely 3D scene based on calibrating Laser scanner and CMOS camera and doing segmentation to result objects in form of region of interest. As a result, the characteristics of each region are then expressed through their corresponding feature vectors, including 2D and 3D features. This is the first time the term of feature vector used to describe a 3D object respecting to the analysis of 3D-point clouds given by a Ladar. Finally, we also prove that the proposed approach leads to more robust and faster processing and decision-making in terrain classification compared with conventional approaches or pixel-based approaches.

Sensor-Fusion Based Real-Time 3D Outdoor Scene Reconstruction and Analysis on a Moving Mobile Outdoor Robot

This article presents an overview over the whole process of generating a precise and rich 3D representation of the local environment of a moving mobile outdoor robot. The resulting model of this process is a camera image textured triangle mesh which is triangulated from a motion-corrected laser scanned 3D point cloud. The demanding requirements of autonomous off-road environment model acquisition are handled by applying a multi-sensor fusion approach. Additionally to the model creation process a novel way of detecting and describing feature points on a piece-wise linear 3D surfaces is presented. The set of feature points generated by the proposed method is a valuable abstraction of a whole outdoor scene that can be used in several following processing steps like mesh simplification or segmentation and robotics-specific tasks like obstacle detection or classification. All described methods are implemented and successfully used on the award-winning robot AMOR.

Distance measuring using calibrating subpixel distances of stereo pixel pairs in artificial compound eye

The electronic cluster eye (eCley) is a new artificial compound eye with some prominent properties of high resolution, miniature volume, and large field of view. Although having the potential capability of obtaining the depth information, the limitation of the short focal length and tiny volume of the eCley still confine the conventional stereo matching algorithms to determine the distances of objects. In this paper, a new method of measuring the subpixel distances of stereo pixel pairs for the eCley is introduced to implement the function of perceiving the depth information. According to the property of the intensity transitional area, the proposed algorithm employs the corrected coefficients to eliminate the deviation of lightness, identifies the transitional area between the objects and background, fits the edge intensity distribution with the Sigmoid function, and derives the optimal subpixel distances of pixel pairs. The detailed experiments and analysis further demonstrate the effectivity of this method. In the real application the method achieves the accuracy of 90% and satisfies real demands.

Development of a High Speed 3D Laser Measurement System for Outdoor Robotics

This paper describes the development of a 3D laser measurement system for application in outdoor robotics. Over the last few years several approaches have been examined concerning this issue. In the majority of cases, a 2D laser scanner, including its whole weight, is used by a rotating motion to obtain a three dimensional model of the near and far environment. To prevent the disadvantages, like particularly the loss of velocity during the scanning process due to the high mass to be accelerated, a novel approach is illustrated within this paper. By mounting a light weight mirror directly in front of the 2D laser scanner, a higher velocity can be achieved while capturing the environment. In addition, further improvements in the minimization of power consumption and also in the reduction considering the complexity of hardware can be reached. The achievement of a 6 Hz frame rate represents the result of this research.