Measurement Of Root Architecture Parameters Based On Digital Image Processing

Plant root architecture, defined the configuration and distribution in the space of root system growing in the soil, has an important effect on plant acquiring nutrients and water from environment substrates such as soil. Research on the characteristic of plant root architecture is useful to investigate the roots distribution, configuration, and function in detail, and estimate the process and ability of roots acquiring nutrient from soil. Virtual plant model is a main methodology for figuring out the relation of plant growth and its growing condition. The difficulty of investigating actual roots makes simulation a very helpful and feasible method as well as a nodus and pivot of roots study. Although quantification of root system by computer simulation of root architecture has academic and applied significance in related research, at present, there is no method to find out modeling parameters of different root architecture for the reason that root system shows many types. Therefore, the scope of application of computer simulation of root architecture is small because all the existing studies on roots simulation is basically restricted to analyzing some specific plant roots.Based on the correlative theory of the growth and configuration of root in plant physiology and plant taxology, this research acquired root static images by digital photography which developed rapidly, and applied digital image processing combined with artifical neural networks technology to root architecture parameters measurement, attempted to explor the method to identify root configuration pattern and measure root architecture parameters. The research work has established the acquisition system of root images, including setting the shooting lamp-house of digital camera and the shooting background etc, and has treat the root images with lower processing, including smoothing by mean filter, and sharpening by fuzzy enhance. Then, high-quality colorful images were acquired for succedent image segmentation.Image segmentation was the key step of image processing. To measure the rootarchitecture parameters, the root system should be picked up from the background as a whole firstly. After multiplicate colorful image segmentation principles were introduced Contrastively, two colorful root image segmentation methods based on basic theory and method of colorful image processing were presented: the method of colorful image segmentation based on color characteristic and the method of colorful image segmentation based on BP neural networks, that could segment the root image well and chang it into a binary image.In order to measure lengths and angles of root accurately, the thesis did experiments on many thining methods, and a serial-parallel thinning method —Frame Reserve Strip (FRS) was applied. This algorithm could thin root image effectively. Especially it could keep the thinning result as single pixel width and single 8-connectivity pixel connection. Also it could keep the image length from shortening basically, and provide accurate skelecton image for root architecture parameters measurement. And a "tracker-marker" method for skelecton image of root system, which could measure the length of tap root, lateral root and lateral root gap, and classify the pixels of root image, was presented;A"cosine-estimated" method and a neural networks method were analysed and applied to measure angles between main root and lateral, the results showed that the BP neural networks method utilized the span of angles as characteristic vectors to measure the angles, which could avoided measurement error caused by the curve of root in the "cosine-estimated" method.Results of experiments showed that the method of measuring root architecture parameters based on digital image processing, presented in the thesis ,could measure root architecture parameters such as length, the angles between main root and lateral roots, number of tips and branching points, and lateral root gap etc. It is the foundation of simulation of root architecture, correlative research and application of root system.