Location: Healthy Processed Foods Research2011 Annual Report
1a. Objectives (from AD-416)
The overall objective is to develop high throughput (~75 samples/s), economically feasible sorting devices for specialty crop product streams. The means for detection will include sensors and/or imaging coupled with algorithm development that differentiates good product from defective, contaminated, or otherwise undesirable product. Sorting devices based on these techniques may be high-speed and non-destructive for mass inspection or as an aid in the sampling and grading process. Specific objectives over the period covered by this project plan are: Detect insect damage in almonds; Detect fungal infestation in walnuts; Detect Olive fly infestation in olives and; Sort shells and kernels in the pistachio nut process stream.
1b. Approach (from AD-416)
1) Appropriate methodologies will be developed to create samples of the product defect to be identified in sufficient quantity for sorter and algorithm development. 2) Using primarily x-ray and color camera imaging technology features will be identified that distinguish undesirable product from good product. 3) Automatic algorithms will be developed to extract the identified features from images in real-time. 4) Material handling systems will be developed to allow construction of prototype sorting devices based on 1 through 3 above. 5) Prototype testing will be conducted to demonstrate the feasibility of implementing the technology in processing plant environments.
3. Progress Report
Sufficient images have been generated for each objective to begin algorithm development. A high speed color camera based sorting instrument has been acquired for rapid collection of larger databases of images as well as implementation of developed sorting devices.
1. Classification of hyperspectral images. Scientists at the National Food Research Institute in Japan have been investigating the use of hyperspectral imaging to detect fruit fly infestations in mangoes but have been unsuccessful in classification efforts based on traditional chemometric procedures. ARS researchers in Albany, CA, have developed classification algorithms that allow discrimination of the infested fruit. These algorithms will allow the rapid discrimination between fruits that are infested by fruit flies and those that are not. When incorporated into an online hyperspectral imaging inspection system, fruits that are infested can be removed from the processing stream. Thus, only uninfested fruits will be selected. This will reduce the need for quarantine treatments such as vapor heat treatment for exported commodities.
Saranwong, S., Haff, R.P., Thanapase, W., Janhiran, A., Kasemsumran, S., Kawano, S. 2011. Feasibility study of utilizing simplified near infrared imaging for detecting fruit fly larvae in intact fruit. Near Infrared Spectroscopy Journal. 19:55-60.