Optical range-finding from image focus

Scope of the study. Much of the labor-intensive work in agriculture consists of reaching out, grasping an object, and then placing the object in a desired position. This repetitious work exploits the unsurpassed hand-eye coordination in human beings. Substitution of machines for manual labor will require simulation of human hand-eye coordination. Most robots in agricultural applications will need the ability to recognize and manipulate three-dimensional objects. With present technology, this requirement makes agricultural robotic systems uneconomical, except for special applications.;A method for gauging the distance from a video camera to an object of interest was investigated. By using a calibrated camera-lens system, range was related to focus. Optimum focus of the image was determined by maximizing the high-frequency content of the Fourier transform of the object image. The Walsh-Hadamard transform was investigated as an alternative focusing function. Software was developed to determine optimum image focus and control a motorized camera lens.;Findings and conclusions. Range values from the video camera to target objects were calculated by the system. Calculated values were compared with measured distances. Differences between calculated and actual distance averaged less than 0.5%. The Walsh-Hadamard transform provided focus information comparable to the Fourier transform. Using double precision floating-point arithmetic, the Walsh-Hadamard transform executed more than three times faster than the Fourier transform. Distance values calculated using the Walsh-Hadamard transform differed from values calculated with the Fourier transform by less than 1%.;This system used a passive, non-triangulation technique to obtain the distance from the machine vision camera to the object of interest. A passive non-triangulation system was the simplest image acquisition requirements, since it does not require a second camera, structured lighting, camera movement, or time-of-flight measurement equipment. This simplicity must be balanced against the limited information obtained from a monocular view of the scene. For range measurement, this method should provide adequate information for robot arm guidance. The simple hardware requirements are also an advantage when designing a rugged, cost-effective system for use in agricultural applications.