Color computer vision for characterization of corn germplasm

Panigrahi, Suranjan
Journal Title
Journal ISSN
Volume Title
Research Projects
Organizational Units
Journal Issue

A color computer vision system was developed at Iowa State University, Ames, Iowa for morphological characterization of corn germplasm. The system consists of a color camera, a PC-AT host computer, a color frame digitizer, a video display monitor, a color video decoder and encoder, and a specially designed lighting chamber. The lighting chamber was specially designed and fabricated to provide uniform lighting for acquiring images of ear corn. The components of the system were matched and interfaced to configure the entire system. A study was conducted to calibrate each component and the entire system to ensure proper functioning of the system components and the acquisition of quality images. Images can be acquired in RGB (Red, Green and Blue) or HSI (Hue, Saturation, and Intensity) color coordinates. The system can provide a maximum resolution of 480 rows x 512 columns x 8 bits per pixels;Ostu's method of automatic thresholding technique was modified to segment the background of the color image of the ear corn. Algorithms and software were developed to extract the boundary of the ear corn image, and to determine the maximum length, maximum width, area and the perimeter of the image;Fractal geometry, moment invariant and knowledge based heuristic approaches were used to classify the shape of the images of ears of corn into one of the four possible shape classes as defined by the International Board of Plant Genetic Resources. These four shape classes are (1) round, (2) cylindrical, (3) conical, and (4) cylindrical-conical. Empirical relations were developed for two fractal based features, i.e. fractal-shape-factor and fractal perimeter to extract shape feature information. Seven higher order moment invariants were computed to represent shape features of the ear corn image in the moment invariant approach. The knowledge based heuristic approach provided the most accuracy of 96% in shape classification on randomly selected 80 ears of corn;Software were developed to define different colors in numeric ranges of hue and saturation values. A rule based expert system was developed to classify the ear corn image into one of the seven color groups based on the colors of the kernels. The software also provides the user an option to determine the average color of the exposed cob on the ear corn image.

Agricultural and biosystems engineering, Agricultural engineering