Chris J. Gliever

Vision-Based 3D Peach Tree Reconstruction for Automated Blossom Thinning

This paper presents research using a correlation-based stereo vision approach to 3D blossom mapping for automated thinning of peach blossoms on perpendicular “V” architecture trees. To this end, a calibrated camera system is proposed, based upon three synchronized ten megapixel cameras and flash illumination for nighttime image acquisition. A correlation-based stereo algorithm, suitable for parallel processing, is developed with the actual scene structure in mind using multiple camera pairs for validating 3D locations, three different certainty metrics, and does not require extrinsic rectification of the images. Results show that mapping accuracy of less than half of a blossom width ( ~ 1 cm) is feasible, and validates the approach as the sensor part of an automated selective blossom thinning system. Furthermore, the effects of the different certainty metrics are examined. They effectively improve the accuracy of blossom positions when the visibility of blossoms is good by removing insecure matches and through qualified selection of subsets of cameras for 3D triangulation. The proposed algorithm is compared and found superior to a popular global optimization algorithm, designed to perform well in another scene structure, demonstrating the quality of correlation-based stereo in practical applications.

Active optical sensors for tree stem detection and classification in nurseries.

Active optical sensing (LIDAR and light curtain transmission) devices mounted on a mobile platform can correctly detect, localize, and classify trees. To conduct an evaluation and comparison of the different sensors, an optical encoder wheel was used for vehicle odometry and provided a measurement of the linear displacement of the prototype vehicle along a row of tree seedlings as a reference for each recorded sensor measurement. The field trials were conducted in a juvenile tree nursery with one-year-old grafted almond trees at Sierra Gold Nurseries, Yuba City, CA, United States. Through these tests and subsequent data processing, each sensor was individually evaluated to characterize their reliability, as well as their advantages and disadvantages for the proposed task. Test results indicated that 95.7% and 99.48% of the trees were successfully detected with the LIDAR and light curtain sensors, respectively. LIDAR correctly classified, between alive or dead tree states at a 93.75% success rate compared to 94.16% for the light curtain sensor. These results can help system designers select the most reliable sensor for the accurate detection and localization of each tree in a nursery, which might allow labor-intensive tasks, such as weeding, to be automated without damaging crops.

Design of a Soil Cutting Resistance Sensor for Application in Site-Specific Tillage

One objective of precision agriculture is to provide accurate information about soil and crop properties to optimize the management of agricultural inputs to meet site-specific needs. This paper describes the development of a sensor equipped with RTK-GPS technology that continuously and efficiently measures soil cutting resistance at various depths while traversing the field. Laboratory and preliminary field tests verified the accuracy of this prototype soil strength sensor. The data obtained using a hand-operated soil cone penetrometer was used to evaluate this field soil compaction depth profile sensor. To date, this sensor has only been tested in one field under one gravimetric water content condition. This field test revealed that the relationships between the soil strength profile sensor (SSPS) cutting force and soil cone index values are assumed to be quadratic for the various depths considered: 0–10, 10–20 and 20–30 cm (r2 = 0.58, 0.45 and 0.54, respectively). Soil resistance contour maps illustrated its practical value. The developed sensor provides accurate, timely and affordable information on soil properties to optimize resources and improve agricultural economy.

Stereo vision blossom mapping for automated thinning in peach

Currently, manual labor is used in the thinning of fruit for ensuring a high yield of marketable fruits. This paper presents ongoing research using a correlation-based stereo vision approach to automated thinning of peach blossoms on perpendicular ‘V’ architecture trees. To this end a calibrated camera system has been designed using three synchronized ten Mpixel cameras and flash illumination. A total station was used to establish blossom ground-truth location. A correlation-based stereo algorithm was developed that is suitable for parallel processing, using multiple camera pairs for validating the correspondence in 3D space, and does not require rectification of the images. The results showed accuracy of less than half of a blossom width (∼ 1cm), provided a good starting point for further development of the algorithm and validated the approach for automated selective blossom thinning application.

Development of an RTK GPS plant mapping system for transplanted vegetable crops.

This study investigated the feasibility of using real-time kinematics (RTK) GPS to automatically map the locations of tomato transplants in the field as they are planted using a vegetable crop transplanter retrofitted with an RTK GPS receiver, and an on board real-time controller. Two detection methods were evaluated for sensing plant location during planting. One method used an infrared light beam sensor to detect the stem location of each plant immediately after planting. The second method used an absolute shaft encoder mounted on the planting wheel to sense the location that each plant was placed in the soil. Odometry was used to determine the actual Easting and Northing GPS coordinates of each plant by interpolation from the original RTK GPS data stream. A field test was conducted to compare the accuracy of this transplant map with actual plant location. The average absolute differences between the automatically generated transplant map and the plant location determined by GPS survey was 0.8 to 2.1 cm in the Northing direction and 1.6 to 3.8 cm in the Easting direction, which was also the travel direction. Results suggest the feasibility of creating an accurate plant map using an RTK GPS equipped transplanter.