Forest Canopy Image Segmentation Algorithm Based On Adaptive Threshold

Forest canopy parameters have been widely used in forest ecosystem research.The parameters of the canopy structure can directly reflect the growth capacity of the vegetation.These parameters provide great help for the measurement and calculation of forest oxygen release rate,carbon sequestration efficiency,and soil and water conservation ability,and they also provide various indicators for the forest ecosystem.Estimated measurements provide an important basis.Forest canopy image segmentation is a key step in using digital image processing methods to obtain forest canopy parameters.The main purpose of the segmentation is to segment the pixels in the sky and vegetation areas in the image so that subsequent canopy parameters can be calculated.The selection of the image segmentation algorithm directly affects the accuracy and real-time performance of segmentation.This paper is based on the National Natural Science Foundation project "Study on Forest Canopy Leaf Area Index Measurement Method Based on Feedback Verification Mechanism"(31370710).It focuses on analyzing and improving the global adaptive threshold and local adaptive threshold algorithm.The main research work is as follows:First,the pretreatment process of forest canopy images was studied.First,select a higher contrast channel image to improve segmentation.Secondly,in order to reduce the error rate of the algorithm in the effective area of traditional fish-eye image,an improved algorithm combining row-by-row scan-by-row method and region growing method is proposed.Second,the contrast is improved by image histogram transformation.Finally,for the problem of forest fisheye canopy image segmentation affected by the phenomenon of reflection or halo phenomenon such as treetops,trunks,etc.,a dereflection algorithm is proposed to reduce the effect of strong light by adjusting the I value of S channel.Secondly,the forest canopy image segmentation algorithm based on global threshold Otsu algorithm is studied.Firstly,the traditional one-dimensional,two-dimensional and three-dimensional Otsu methods are analyzed and their deficiencies are analyzed.For the low-dimensional Otsu processing,there is a big error.The three-dimensional Otsu method has too slow speed.The method of reducing the dimensionality of the three-dimensional Otsu algorithm is used to differentiate the three dimensions.For 1D and 2D Otsu algorithms.Secondly,for the problem that the decomposed two-dimensional Otsu anti-noise performance is weak,the curve threshold is selected and the concept of intra-class similarity is proposed.Finally,for the two-dimensional Otsu after decomposition,there still exists the problem of long optimization time.The advantages and disadvantages of traditional particle swarm optimization(PSO)and firefly algorithm(FA)are analyzed.The improved FA method is used for optimization.Then,the forest canopy image segmentation algorithm based on local adaptive threshold is studied.First,for the blocky effect problem,the sub-image shape is changed to the fan-shaped region under the same azimuth and zenith angle range,and divided by the Otsu algorithm.Secondly,in order to further improve the block effect phenomenon,the Bemsen algorithm based on pixel neighborhood sliding threshold is used for processing,and Bernsen's threshold rules are improved to realize the automatic and effective region extraction process.Secondly,in order to further speed up the segmentation speed and improve the segmentation effect,the Niblack algorithm is used to implement the segmentation process,and the shortcomings of the traditional algorithms are analyzed.The method of increasing the minimum standard deviation threshold is proposed to reduce the over-segmentation of the algorithm.Finally,a forest canopy image analysis system was designed.The hardware acquisition system adopts a three-axis self-balancing intelligent head with the SToRM32-BGC gyroscope acceleration controller as the core,and incorporates a fish-eye camera;the software analysis system adopts the MATLAB GUIDE development environment design user interface,which can achieve efficient processing on the PC side.Forest fisheye canopy image.