Spatial and Spectral Methods for Weed Detection and Localization
- PDF / 4,145,318 Bytes
- 7 Pages / 600 x 792 pts Page_size
- 83 Downloads / 186 Views
Spatial and Spectral Methods for Weed Detection and Localization Jean-Baptiste Vioix UMR CPAP ENESAD CEMAGREF 21, boulevard Olivier de Serres, 21800 Quetigny, France Email: [email protected]
Jean-Paul Douzals UMR CPAP ENESAD CEMAGREF 21, boulevard Olivier de Serres, 21800 Quetigny, France Email: [email protected]
´ eric ´ Truchetet Fred Le2i, IUT Le Creusot 12, rue de la Fonderie, 71200 Le Creusot, France Email: [email protected]
´ Louis Assemat INRA, Unit´e de Malherbologie et Agronomie, BP 86510 21065 Dijon Cedex, France Email: [email protected]
Jean-Philippe Guillemin ENESAD laboratoire CBF, 21 boulevard Olivier de Serres, 21800 Quetigny, France Email: [email protected] Received 26 July 2001 and in revised form 7 February 2002 This study concerns the detection and localization of weed patches in order to improve the knowledge on weed-crop competition. A remote control aircraft provided with a camera allowed to obtain low cost and repetitive information. Different processings were involved to detect weed patches using spatial then spectral methods. First, a shift of colorimetric base allowed to separate the soil and plant pixels. Then, a specific algorithm including Gabor filter was applied to detect crop rows on the vegetation image. Weed patches were then deduced from the comparison of vegetation and crop images. Finally, the development of a multispectral acquisition device is introduced. First results for the discrimination of weeds and crops using the spectral properties are shown from laboratory tests. Application of neural networks were mostly studied. Keywords and phrases: weed detection, spatial analysis, spectral analysis, Gabor filter, neural network, image processing.
1.
INTRODUCTION
Weed detection is extensively studied, as herbicide application has a relevant impact on farm economics and environment. Developing a spraying strategy in the context of precision agriculture needs to improve in-field detection of weeds. According to literature, weed detection using image analysis was directed through different approaches. First experimental works were based on the spectral signature of weeds and crops. Vrindts and de Baerdemaeker [1] determined some specific spectral bands to achieve weed identification. Statistical analyses were conducted to find spectral properties of each species. In the same way Pollet et al. [2] developed an imaging spectrograph. This device gave an image with
the spatial dimension on vertical axis and the spectral dimension on horizontal axis. Another experimental method was based on morphological properties extracted from leaf shape using simple geometric shape factors (elongation, diameter, etc.) [3, 4]. In the same way Manh et al. used deformable templates to modelize leaf shape [5, 6]. In these two last cases, high resolution images were needed. Moreover, computation time was very important and limited that last application to small area investigation. In-field detection of weeds was also possible on stubble. For example, Biller et al. [7]
Data Loading...
