Machine Vision Spots Spoiled Wheat

Michael D. Wheeler

In the past two years, wheat farmers in the US have faced a serious problem: Fungal spores, known as karnal bunt, have been infecting their crops. To prevent the spread of this blight, farmers and scientists have turned to photonics to speed up detection of the spores and determine whether certain crops are suitable for export.

While wheat infected with karnal bunt is not deadly, it turns kernels dark and gives the wheat a foul, "fishy" smell. Although infected wheat is suitable for livestock feed, it is unfit for human consumption. Contaminated wheat also fetches a much lower price on the world market.

Photonics is helping armers certify that their wheat is free of fungal spores.

Scientists in Arizona, New Mexico and Texas began widespread efforts to designate which wheat crops contained the karnal bunt spores. The process was painstaking and labor-intensive, involving filtering, washing and placing samples of the wash on a microscope slide. Technicians spent more than an hour and a half to inspect each slide for the tiny spores, and they still had an error rate of 20 percent.

Quick detection

Fearing a spread of the spores into Kansas -- the top wheat-producing state in the US -- the Kansas Department of Agriculture contacted AlliedSignal Federal Manufacturing & Technologies in Albuquerque. With money from the US Department of Energy's Technical Assistance Program, AlliedSignal proposed a solution: machine vision.

The company custom-built a system from mostly off-the-shelf technology. Engineers mounted a black-and-white charge-coupled device video camera manufactured by Sony to a Nikon Labophot-2 microscope for image acquisition. It built a computer-controlled stage for slide positioning and installed a video frame grabber card manufactured by Alacron Inc. in Nashua, N.H., in a 166 Pentium PC. This enabled the acquisition of each 480 × 640-pixel image. Specially designed software performed pattern recognition and feature extraction.

Chris Baumgart, a senior technology manager at AlliedSignal, said since Kansas has started using the system, evaluation time has decreased to about 15 minutes per slide. The system also boasts a dramatically lower error rate. This has enabled farmers to evaluate 100 acres of wheat in the time it previously took to inspect 10.

In addition, farmers looking to export their wheat to countries such as China can certify their wheat is free of karnal bunt. Baumgart said that AlliedSignal has investigated other applications for its machine vision system, including cases of corn blight and other crop diseases.