Location: Application Technology ResearchTitle: Robust crop and weed segmentation under uncontrolled outdoor illumination) Author
|Jeon, Hong Young|
Submitted to: Sensors
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 6/7/2011
Publication Date: 7/14/2011
Publication URL: http://hdl.handle.net/10113/50088
Citation: Jeon, H., Tian, L.F., Zhu, H. 2011. Robust crop and weed segmentation under uncontrolled outdoor illumination. Sensors. 11(6)6270-6283. Interpretive Summary: Existing machine visions used for precision weed-control sprayers require manual image processes to increase their weed detection accuracy. However, the manual processes hamper to automatically apply herbicides to targets during the spray application. An automated algorithm to identify plants and detect weeds was developed for the machine vision technology. The algorithm was implemented into a 4-wheel autonomous research robot and was tested with images captured under various field illumination conditions. Test results demonstrated that the new algorithm was able to identify plants and detect weeds steadily in the field. Consequently, this algorithm ensured a reliable weed detection technique for the development of robotic sprayers to automatically apply herbicides for weed controls.
Technical Abstract: A new machine vision for weed detection was developed from RGB color model images. Processes included in the algorithm for the detection were excessive green conversion, threshold value computation by statistical analysis, adaptive image segmentation by adjusting the threshold value, median filter, morphological feature calculation and Artificial Neural Network (ANN). The developed algorithm was validated for its ability to identify and detect weeds and crop plants. A field robot implementing the machine vision was tested in the field to capture field images under various outdoor illumination conditions. The error of the algorithm to process 666 field images ranged from 2.1 to 2.9 %. The ANN correctly detected 72.6 % of crop plants from the identified plants, and considered the rest as weeds. However, the ANN detection rates for crop plants were improved up to 95.1 % by coordinating the error sources in the algorithm. The developed weed detection and image processing algorithm provides a novel method to identify plants against soil background under the uncontrolled outdoor illuminations, and to defer weeds from the plants. Thus, the new machine vision may be useful for outdoor applications including plant specific direct applications (PSDA).