Research On Compressive Sensing For Plant Hyperspectral Data

Due to the combination of image and spectral information,hyperspectral remote sensing technology has been widely used in agriculture in recent years.However,with the pursuit of higher spatial resolution and spectral resolution,hyperspectral image data with a large amount of original data is more difficult to store and transmit.The emergence of compressive sensing technology has alleviated this problem.In recent years,some compressive sensing techniques for plant hyperspectral have been proposed.However,these techniques focus on exploring how to use hyperspectral spatial correlation and spectral correlation to obtain a priori information to aid compressive sensing to improve the quality of reconstruction,ignoring improving compressive sensing technology from the perspective of plant hyperspectral practical application requirements.The specific research contents of this paper include:(1)Aiming at the problem of significant noise in some bands of plant hyperspectral,this paper explores how to remove significant noise in the process of compressive sensing,and proposes a plant hyperspectral compressive sensing technology based on linear prediction model(PSSAHCS).(2)The practical application of plant hyperspectral is often only concerned with plant areas,and the existing plant hyperspectral compressive sensing technology does not distinguish between plant areas and background areas.Compressive sensing techniques are researched at leaf scale and canopy scale respectively.(3)At the leaf scale,the hyperspectral compressive sensing technology of plant leaves based on image segmentation technology was studied(ISABCS),and the real hyperspectral data of tea was used for experiments.(4)At the canopy scale,the plant hyperspectral canopy segmentation technique based on joint sparse representation classification was studied(JSCCS),and then the real canopy hyperspectral data was used to quantitatively analyze the key parameters involved in the technology.(5)The tensor compressive sensing technology for plant hyperspectral canopy regions is proposed(JATCS),and the internal structure is preserved under the premise of ensuring the quality of canopy reconstruction.Experiments were performed using real canopy hyperspectral data.The experimental results show that PSSAHCS is able to effectively remove the significant noise in some bands during the process of compressive sensing.ISABCS can realize compressive sensing only on the leaf's regions,and the quality of the spatial and spectral domain reconstruction in the leaf's regions is higher.JSCCS can achieve accurate segmentation of the canopy region.JATCS can realize tensor compressive sensing only on the canopy region,and the reconstructed canopy region not only has higher spatial and spectral quality,but also retains original structure completely.