This paper investigates the potential of a consumer-grade infrared stereo camera, i.e. the Intel RealSense D435, to automatically extract crop status information, such as Normalized Difference Vegetation Index (NDVI), in arable and permanent crops. The sensing device includes two infrared (IR) sensors for depth calculation and one colour sensor, which provide, for each point of the scene, both IR and visible light information thus making it possible pixel per pixel NDVI estimations. Measurements were performed on various arable crops including corn (Zea mays) and barley (Ordeum vulgare) and on two vine varieties, Freisa and Malvasia, and were compared to measurements taken by a Trimble GreenSeeker handheld crop sensor. Results show that the RealSense camera tends to underestimate NDVI values compared to the GreenSeeker, with squared correlation coefficient r2 = 0.68. The fitted regression equation is successively applied to correct new camera observations, resulting in good agreement with the GreenSeeker output. The use of the RGB-D camera to simultaneously provide canopy height measurements by a farmer robot is also demonstrated in a Malvasia field, showing that the proposed system can be effectively adopted for fully automated plant-scale monitoring of vineyards.
This paper describes a novel image acquisition and processing framework to detect and count grape bunches along vineyard rows, using RGB and depth data acquired in the field by a farmer robot. The proposed pipeline starts with a semantic image segmentation module that uses a pre-trained convolutional neural network to separate fruit from non-fruit regions. Areas pertaining to fruits are then further processed using a depth gradient-based clustering algorithm to detect and separate single grape bunches. Experiments performed in a commercial vineyard are presented showing that, despite the low quality of the input images, the proposed approach is able to correctly detect and count grape clusters with good accuracy.
Wheel-terrain interaction plays a critical role for vehicle mobility on natural terrain, such as in agricultural, planetary exploration and off-road settings. Estimation of the terrain characteristics and the way they affect traversability is essential for the vehicle to better plan its safest and energy-efficient path. This work proposes a novel approach to learn and predict from a distance the motion resistance encountered by a robotic vehicle, while traversing natural soil, by using visual information from a stereovision device. To this end, terrain appearance and geometry information are first correlated to resistance torque measurements during a learning phase via two alternative regression approaches, namely Least-Squares Boosting and Long-Short Term Memory Recurrent Neural Network. Then, such a relationship is exploited to predict motion resistance remotely, based on visual data only. Results obtained in preliminary experimental tests on ploughed and compact terrain are presented to show the feasibility of the proposed method.
Global navigation satellite system (GNSS) is the standard solution for solving the localization problem in outdoor environments, but its signal might be lost when driving in dense urban areas or in the presence of heavy vegetation or overhanging canopies. Hence, there is a need for alternative or complementary localization methods for autonomous driving. In recent years, exteroceptive sensors have gained much attention due to significant improvements in accuracy and cost-effectiveness, especially for 3D range sensors. By registering two successive 3D scans, known as scan matching, it is possible to estimate the pose of a vehicle. This work aims to provide in-depth analysis and comparison of the state-of-the-art 3D scan matching approaches as a solution to the localization problem of autonomous vehicles. Eight techniques (deterministic and probabilistic) are investigated: iterative closest point (with three different embodiments), normal distribution transform, coherent point drift, Gaussian mixture model, support vector-parametrized Gaussian mixture and the particle filter implementation. They are demonstrated in long path trials in both urban and agricultural environments and compared in terms of accuracy and consistency. On the one hand, most of the techniques can be successfully used in urban scenarios with the probabilistic approaches that show the best accuracy. On the other hand, agricultural settings have proved to be more challenging with significant errors even in short distance trials due to the presence of featureless natural objects. The results and discussion of this work will provide a guide for selecting the most suitable method and will encourage building of improvements on the identified limitations.
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.