Weed detection in lawn field based on gray-scale uniformity

摘要

It is necessary that the weed area is discriminated from lawn area if selective spraying is applied to maintain a lawn field. However, both weed and lawn usually have similar green color in summer. In this paper, a color independent method using gray-scale uniformity in image was proposed for detecting the weed area. This method works based on different levels of uniformity for weed and lawn surfaces and could precisely detect only the weed leaves. The analysis results showed that a weed could be detected if its leaves are at least 5 pixels wide in image, while its length is enough to .generate a minimum size blob. It was also found that the image had to be acquired at a short camera distance so that the soil among the lawn was visible and the lawn area had less uniformity. After weed leaf detection, a morphological image analysis was performed on a group the adjacent blobs as result of leaf detection. It was observed that the morphological image analysis was useful in locating a representative center-point for grouped blobs in correspondence to main root location of the weeds.