C. T. Morrow

Machine vision for color inspection of potatoes and apples

A machine vision system was trained to distinguish between good and greened potatoes and yellow and green ‘Golden Delicious’ apples. The method of using the HSI (Hue, Saturation, and Intensity) color system proved highly effective for color evaluation and image processing. The vision system achieved over 90% accuracy in inspection of potatoes and apples by representing features with hue histograms and applying multivariate discriminant techniques. Reducing the number of hue bins by selecting significant features only or by summing groups of hue bins increased misclassification by the vision system. Color classification represents an important quality feature evaluation method that needs to be integrated into an overall automated quality inspection and grading system.

Fourier-based Separation Technique for Shape Grading of Potatoes Using Machine Vision

A Fourier-based shape separation method was developed for shape grading of potatoes using machine vision for automated inspection. The relationship between object shape and its boundary spectrum values in Fourier domain was explored for shape extraction. A new and fast method of using Green’s theorem and boundary Fourier coefficients was given for estimating elongation of an object. A shape separator based on harmonics of the transform was defined for potato shape separation. Tests showed the shape separator was effective and efficient for difficult shape separation. The machine vision system developed has a great potential to assist humans for automated potato grading.

Grading of Mushrooms Using a Machine Vision System

The quality features of the common white Agaricus bisporus mushroom were quantified using image analysis in order to inspect and grade the mushrooms by an automated system. The features considered were color, shape, stem cut, and cap veil opening. Two human inspectors evaluated samples which were divided into training and test sets. The vision system was trained to classify mushrooms into two quality grades using thresholding. The human inspection results were compared with each other as well as the computer vision system results. Misclassification by the vision system ranged from 8 to 56% depending upon the quality feature evaluated, but averaged about 20%. The disagreement between inspectors ranged from 14 to 36%.

Machine Vision Inspection of ‘Golden Delicious’ Apples

A machine vision system was developed to form a basis for single-pass quality feature inspection and grading of ‘Golden Delicious’ apples. The inspection criteria were based on USDA standards for fresh market apples. Image analysis algorithms were developed to assess and quantify the quality features of color, shape, and russet. Over 300 ‘Golden Delicious’ apples were inspected by the machine vision system and the results were compared to a human inspector. The vision system was able to correctly classify 100% of the apples for color, 92.3% for shape, and 82.5% for russetting.

Comparison of a Neural Network and Traditional Classifier for Machine Vision Inspection of Potatoes

This work addressed the relative strengths and weaknesses of the backpropagation neural network versus the Fisher discriminant function. Their performance was compared for machine vision inspection of greening, shape, and shatter bruise in two potato cultivars. The backpropagation network’s number of hidden nodes were varied from zero to eight for each defect type to determine the optimal network classification size. The network was trained and tested five times at each hidden node number and defect type to minimize local minima variation. For greening, the best backpropagation network averaged 74.0% with three hidden nodes while the Fisher method performed with a 70.0% accuracy. The backpropagation method also performed better for shape discrimination with a 73.3% average accuracy at seven hidden layer nodes versus a 68.1% accuracy. The Fisher method performed better for shatter bruise detection with a 76.7% accuracy versus a 56.0% average accuracy at four hidden layer nodes for backpropagation.

IDENTIFICATION AND SEPARATION OF MICROPROPAGATED SUGARCANE SHOOTS BASED ON THE HOUGH TRANSFORM

A Hough transform-based computer algorithm was developed to identify sugarcane shoots as straight lines in a micropropagated Stage 2 sugarcane shoot clump image. An arc-based shoot identification method developed previously was tested after changing the lighting from fluorescent to incandescent which improved contrast between stems and leaves. This method was then compared to the Hough transform method. The Hough transform algorithm correctly identified 93% of total shoots versus 82% for the arc method, missed 7% of the shoots versus 18% for the arc method, and improperly identified non-shoots as shoots 6% of the time versus 7% for the arc method.

AUTOMATED CYCLED SPRINKLER IRRIGATION SYSTEM FOR FROST PROTECTION OF APPLE BUDS

An automated cycled overtree sprinkler irrigation system was implemented and tested in a 0.4 ha (1 acre) dwarf apple orchard to protect apple buds from cold temperatures. The system reduced water usage compared to a more conventional approach of continuous sprinkling. The control scheme was based on a system that monitored the environmental parameters (air temperature, wind speed, relative humidity) and bud temperatures, calculated the on and off times, and cycled the valve. The system was tested during three frost events in the spring of 1997. The control system successfully kept the bud temperatures above the critical level during two of three frost events. During one event the sprinkled orchard temperatures dropped below the critical temperature for a short duration but were warmer than the unsprinkled orchard temperatures. The average reduction in water during the three frost events tested was about 72% as compared to continuous water application using the same system.

Automation of an Intermittent Overhead Irrigation Frost Protection System for an Apple Orchard

An automated sprinkling frost protection system was developed and tested to reduce the environmental damage caused by excess irrigation. A computer algorithm made the initial “turn-on” decision, cycled the irrigation, and turned off the system when the frost event ended. The sprinkling turn-on decision is based on bud stage of development, air temperature, wind speed, relative humidity, and rate of temperature fall. Sprinkler on-off cycles were used to adjust for a recommended sprinkler application rate determined by an energy balance model. The automated system was tested on an apple orchard during one spring and two fall frost events in 1992 and two spring frost events in 1993. The system successfully kept the apple flower bud temperatures above the critical bud temperature during two of the four frost events, while the bud temperatures dropped below the critical values for short durations on the other two events. The system has the potential to save water as indicated by a water use reduction of 75% during one mild frost event compared with the conventional approach of continuous sprinkling.

Improving Efficiency in a Specialty Baked Product Line

A systems analysis was performed on a baked apple product line in a fruit processing plant. The purpose of the analysis was to determine the energy usage of the processing line and to improve productivity through automation. The energy usage audit focused on the baking oven within the line, since the oven consumed more energy than other component. The possibility of replacing the gas oven with a microwave oven was explored. Results of the audit show that although the oven is 30 years old, it is reasonably energy efficient and replacing it would not be cost-effective. Electrical energy usage was 0.192 MJ/kg (82.54 Btu/lb) of baked product and gas energy usage was 0.139 MJ/kg (59.75 Btu/lb) of baked product, neither of which was considered to be excessive. The biggest need identified within the plant was for automation since the company currently faces a decreasing labor pool. Several suggestions for automating line components were provided.