Click on “Download PDF” for the PDF version or on the title for the HTML version.

If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.

Abstract. Despite modern advances in computer vision technologies, there is no off-the-shelf real-time weed detection system to integrate with Unmanned Aerial System (UAS) for real-time spot spraying of weed. The primary challenge is the lack of available fast weed detection algorithms deployable in lightweight, robust, and inexpensive computers like Raspberry Pi which is suitable to use in UAS. One of the most common methods used to find a line, circle, or any parametric feature is the Hough Transform. However, Hough Transform is computationally heavy and unsuitable for real-time weed detection. A faster method is needed to implement in real-time. We developed the Random Point Method (RPM) as a more efficient method to detect lines and circles in an image. The algorithm takes an edge-detected binary image as an input. The image is then divided into small grids of desired resolution. We compared the performance of the Hough Transform and the RPM in the Matlab environment with real images of clover a common weed. When a 3280x2464-pixel image with clover weed was processed using the Hough Transform, it took 2.37 seconds to process the image to find clover leaf centers. RPM method, on the other hand, took only 0.018 seconds to process the same image, which is about 131 times faster than the Hough Transform.