Machine Vision System for Spot‑Application
Currently, crop management practices are implemented uniformly with inadequate attention being given to the weed distribution within field. These variations within wild blueberry fields emphasize the need for precise site-specific crop management. Moreover, the wild blueberry fields have significant bare spots (30-50% of total field area). In these situations, variable rate application of agrochemicals holds great potential to allocate inputs more efficiently by exploiting the locations, timings, and rates of herbicide. Previous machine vision techniques cannot be utilized practically in a mobile spraying system because of high cost cameras (hyper spectral) and the time between image capture and spraying was too short to enable the algorithm to process the crop or field condition. Therefore, a very efficient machine vision algorithm using cost effective digital color camera was developed, that can process the images and underlying field condition to overcome the time delay issues with the previously available systems while the mobile sprayer system advances at typical ground speed.
In general, the present research provides computer-readable media for variable spraying of agrochemicals. The present research uses computationally efficient techniques for detecting the crop or field condition based on digital imagery. Rather than relying on computationally expensive machine vision techniques, the present research uses an innovative green detection technique for image decomposed into RGB components or an innovative textural feature technique that uses only a subset of textural features drawn from luminance, hue, saturation and intensity co-occurrence matrices. These algorithms enable real-time detection of the crop conditions and/or soil or ground conditions so that the variable rate sprayer can dispense an appropriate amount of agrochemical with specific sections of the boom where the targets were detected while the mobile sprayer advances at a normal speed. Research on this project has been on-going since 2009. Currently, the 45 ft. MS 1135E sprayer boom is divided into sixteen sections and mounted behind a John Deere 6430 farm tractor. Sixteen solenoid valves and nozzles are installed on the rear boom. Eight cameras (one for every two sections) are incorporated vertically on a 45 ft front mounted boom. Self leveling sensing system was introduced in front boom to adjust camera heights automatically during field operation. Two 8-channel computer controllers receive triggering signals from the custom-made image processing software and communicate automatically to the Dickey John controller. The Dickey John controller regulates the flow rate to the nozzles through a servo valve.
Dr. Qamar Zaman, Associate Professor, Engineering Department, NSA
Dr. Young Ki Chang (Post Doctoral Fellow)
Dr. Arnold Schumann, Associate Professor, Citrus Research and Education Centre, University of Florida, USA
Travis Esau (Graduate Student)
This research was funded by Oxford Frozen Foods Limited, Agri-Futures (ACAAF) Nova Scotia, Wild Blueberry Producers Association of Nova Scotia and Department of Agriculture Technology Development Program, Canadian Foundation for Innovation, NS Growing Forward, Department of Agriculture.