This paper describes a robot system for the automatic pruning of grape vines. A mobile platform straddles the row of vines, and it images them with trinocular stereo cameras as it moves. A computer vision system builds a three-dimensional (3D) model of the vines, an artificial intelligence (AI) system decides which canes to prune, and a six degree-of-freedom robot arm makes the required cuts. The system is demonstrated cutting vines in the vineyard. The main contributions of this paper are the computer vision system that builds 3D vine models, and the test of the complete-integrated system. The vine models capture the structure of the plants so that the AI system can decide where to prune, and they are accurate enough that the robot arm can reach the required cuts. Vine models are reconstructed by matching features between images, triangulating feature matches to give a 3D model, then optimizing the model and the robot's trajectory jointly (incremental bundle adjustment). Trajectories are estimated online at 0.25 m/s, and they have errors below 1% when modeling a 96 m row of 59 vines. Pruning each vine requires the robot arm to cut an average of 8.4 canes. A collision-free trajectory for the arm is planned in intervals of 1.5 s/vine with a rapidly exploring random tree motion planner. The total time to prune one vine is 2 min in field trials, which is similar to human pruners, and it could be greatly reduced with a faster arm. Trials also show that the long chain of interdependent components limits reliability. A commercially feasible pruning robot should stop and prune each vine in turn. C 2016 Wiley Periodicals, Inc.

No abstract

Abstract-It is common in mobile augmented reality (AR) research to examine results that were attained with unpublished tools and data sets, which makes it difficult to compare and improve existing work without significant effort. We discuss the development of an open source toolkit called Transform Flow, which includes a data capture application for iOS, a desktop application to replay and analyse captured data sets with different algorithms, and a mobile application that can run these algorithms in real-time. Our results suggest that our toolkit can be a centre for collaborative research, as it provides a common platform on which tracking algorithms for mobile AR can be developed, studied and eventually deployed.

No abstract

Detailed within is an attempt to implement a real-time radar signal classification system to monitor and count bee activity at the hive entry. There is interest in keeping records of the productivity of honeybees. Activity at the entrance can be a good measure of overall health and capacity, and a radar-based approach could be cheap, low power, and versatile, beyond other techniques. Fully automated systems would enable simultaneous, large-scale capturing of bee activity patterns from multiple hives, providing vital data for ecological research and business practice improvement. Data from a Doppler radar were gathered from managed beehives on a farm. Recordings were split into 0.4 s windows, and Log Area Ratios (LARs) were computed from the data. Support vector machine models were trained to recognize flight behavior from the LARs, using visual confirmation recorded by a camera. Spectrogram deep learning was also investigated using the same data. Once complete, this process would allow for removing the camera and accurately counting the events by radar-based machine learning alone. Challenging signals from more complex bee flights hindered progress. System accuracy of 70% was achieved, but clutter impacted the overall results requiring intelligent filtering to remove environmental effects from the data.