Yael Edan

Vision-based hand-gesture applications

Body posture and finger pointing are a natural modality for human-machine interaction, but first the system must know what its seeing.

Harvesting Robots for High-value Crops: State-of-the-art Review and Challenges Ahead

This review article analyzes state-of-the-art and future perspectives for harvesting robots in high-value crops. The objectives were to characterize the crop environment relevant for robotic harvesting, to perform a literature review on the state-of-the-art of harvesting robots using quantitative measures, and to reflect on the crop environment and literature review to formulate challenges and directions for future research and development. Harvesting robots were reviewed regarding the crop harvested in a production environment, performance indicators, design process techniques used, hardware design decisions, and algorithm characteristics. On average, localization success was 85%, detachment success was 75%, harvest success was 66%, fruit damage was 5%, peduncle damage was 45%, and cycle time was 33 s. A kiwi harvesting robot achieved the shortest cycle time of 1 s. Moreover, the performance of harvesting robots did not improve in the past three decades, and none of these 50 robots was commercialized. Four future challenges with R&D directions were identified to realize a positive trend in performance and to successfully implement harvesting robots in practice: 1 simplifying the task, 2 enhancing the robot, 3 defining requirements and measuring performance, and 4 considering additional requirements for successful implementation. This review article may provide new directions for future automation projects in high-value crops.

Computer vision for fruit harvesting robots state of the art and challenges ahead

Despite extensive research conducted in machine vision for harvesting robots, practical success in this field of agrobotics is still limited. This article presents a comprehensive review of classical and state-of-the-art machine vision solutions employed in such systems, with special emphasis on the visual cues and machine vision algorithms used. We discuss the advantages and limitations of each approach and we examine these capacities in light of the challenges ahead. We conclude with suggested directions from the general computer vision literature which could assist our research community meet these challenges and bring us closer to the goal of practical selective fruit harvesting robots.

Near-minimum-time task planning for fruit-picking robots

A near-minimum-time task-planning algorithm for fruit-harvesting robots having to pick fruits at N given locations is presented. For the given kinematic and inertial parameters of the manipulator, the algorithm determines the near-optimal sequence of fruit locations through which the arm should pass and finds the near-minimum-time path between these points. The sequence of motions was obtained by solving the traveling salesman problem (TSP) using the distance along the geodesics in the manipulators inertia space, between every two fruit locations, as the cost to be minimized. The proposed algorithm was applied to define the motions of a citrus-picking robot and was tested for a cylindrical robot on fruit position data collected from 20 trees. Significant reduction in the required computing time was achieved by dividing the volume containing the fruits into subvolumes and estimating the geodesic distance rather than calculating it. >

Robotic melon harvesting

Intelligent sensing, planning, and control of a prototype robotic melon harvester is described. The robot consists of a Cartesian manipulator mounted on a mobile platform pulled by a tractor. Black and white image processing is used to detect and locate the melons. Incorporation of knowledge-based rules adapted to the specific melon variety reduces false detections. Task, motion and trajectory planning algorithms and their integration are described. The intelligent control system consists of a distributed blackboard system with autonomous modules for sensing, planning and control. Procedures for evaluating performance of the robot performing in an unstructured and changing environment are described. The robot was tested in the field on two different melon cultivars during two different seasons. Over 85% of the fruit were successfully detected and picked.

Grape clusters and foliage detection algorithms for autonomous selective vineyard sprayer

While much of modern agriculture is based on mass mechanized production, advances in sensing and manipulation technologies may facilitate precision autonomous operations that could improve crop yield and quality while saving energy, reducing manpower, and being environmentally friendly. In this paper, we focus on autonomous spraying in vineyards and present four machine vision algorithms that facilitate selective spraying. In the first set of algorithms we show how statistical measures, learning, and shape matching can be used to detect and localize the grape clusters to guide selected application of hormones to the fruit, but not the foliage. We also present another algorithm for the detection and localization of foliage in order to facilitate precision application of pesticide. All image-processing algorithms were tested on data from movies acquired in vineyards during the growing season of 2008 and their evaluation includes analyses of the potential pesticide and hormone reduction. Results show 90% accuracy of grape cluster detection leading to 30% reduction in the use of pesticides. The database of images is placed on the Internet and available to the public to continue developing the detection algorithms.

Navigation of decentralized autonomous automatic guided vehicles in material handling

This paper presents a navigation methodology for decentralized autonomous automated guided vehicles used for material handling. The navigation methodology is based on behavior-based control augmented with multirobot coordination behaviors and a priori waypoint determination. Results indicate that the developed methodology fuses well between the desires for optimal vehicle routes on the one hand and decentralized reactive operation on the other.

Human‐robot collaboration for improved target recognition of agricultural robots

Automatic target recognition in agricultural harvesting robots is characterized by low detection rates and high false alarm rates due to the unstructured nature of both the environment and the objects. To improve detection human‐robot collaboration levels were defined and implemented. The collaboration level is defined as the level of system autonomy or the level at which the human operator (HO) interacts with the system. Experimental results on images taken in the field indicate that collaboration of HO and robot increases detection and reduces the time required for detection.

IMAGE–PROCESSING ALGORITHMS FOR TOMATO CLASSIFICATION

Image–processing algorithms were developed and implemented to provide the following quality parameters for tomato classification: color, color homogeneity, defects, shape, and stem detection. The vision system consisted of two parts: a bottom vision cell with one camera facing upwards, and an upper vision cell with two cameras viewing the fruit at 60o. The bottom vision cell determined fruit stem and shape. The upper vision cell determined fruit color, defects, and color homogeneity. Experiments resulted in 90% correct bruise classification with 2% severely misclassified; 90% correct color homogeneity classification; 92% correct color detection with 2% severely misclassified, and 100% stem detection.

Cluster labeling and parameter estimation for the automated setup of a hand-gesture recognition system

In this work, we address the issue of reconfigurability of a hand-gesture recognition system. The calibration or setup of the operational parameters of such a system is a time-consuming effort, usually performed by trial and error, and often causing system performance to suffer because of designer impatience. In this work, we suggest a methodology using a neighborhood-search algorithm for tuning system parameters. Thus, the design of hand-gesture recognition systems is transformed into an optimization problem. To test the methodology, we address the difficult problem of simultaneous calibration of the parameters of the image processing/fuzzy C-means (FCM) components of a hand-gesture recognition system. In addition, we proffer a method for supervising the FCM algorithm using linear programming and heuristic labeling. Resulting solutions exhibited fast convergence (in the order of ten iterations) to reach recognition accuracies within several percent of the optimal. Comparative performance testing using three gesture databases (BGU, American Sign Language and Gripsee), and a real-time implementation (Tele-Gest) are reported on.