| As a training suit for military camouflage,camouflage clothing consists of irregular protective color patterns of colors such as green,yellow,tea,and black.The quality of camouflage is directly related to the survival of soldiers and the victory or defeat of war.Color is an important factor in measuring the camouflage effect of camouflage fabrics.The more the color of camouflage fabrics fits the spectral reflectance of the target background environment,the lower the casualty rate of soldiers and the higher the combat effectiveness of the army as a whole.Therefore,color quality detection of camouflage fabrics is very important.Camouflage fabric patterns are designed according to various complicated geographical environments,the problems of different size,shape and arrangement of patterns make the traditional color evaluation or measurement methods have certain limitations.In order to obtain accurate color information of camouflage,a color measurement method of camouflage fabrics based on multi-spectral images is proposed in this paper.By obtaining the pixel point response value information of the image,the spectral reflectance of camouflage fabrics is reconstructed by spectral reconstruction algorithm,and the color measurement of camouflage fabrics is realized.First,the existing methods of fabric color measurement and evaluation are analyzed.Through the actual measurement of camouflage color,it is found that the spectrophotometric color measurement method cannot measure the areas with too complex and too small camouflage blocks,and the measurement efficiency is low because of the contact measurement;The DigiEye image color measurement system can only measure the color within the area.Secondly,a multi-spectral image acquisition system based on LED adjustable light source and digital camera is built in this paper,which can quickly obtain color images of camouflage fabrics and the multi-spectral image under nine narrowband wavelengths(405nm,425 nm,450nm,475 nm,505nm,525 nm,595nm,635 nm,660nm).This paper studies the image processing technology to realize the segmentation of camouflage color blocks,and proposes a segmentation method based on the combination of color image and multi-spectrum.The segmentation results of color images are obtained by using image filtering and image segmentation techniques,and then the position information of pixels is mapped to multi-spectral images.The results show that mean shift clustering can quickly realize the segmentation of multi-spectral images of the same kind.Then this paper studies the spectral reconstruction method based on learning algorithm.Aiming at the influence of different training sets on the spectral reconstruction of camouflage prediction samples,two training sample sets are proposed: the global sample set of DCC standard color plate and the local training sample set based on camouflage color itself,and the training set spectrum by four spectral reconstruction methods.By comparing the results of four spectral reconstruction methods,it is found that the results of violation method and principal component analysis method are far inferior to those of BP neural network regression and support vector machine regression,and support vector machine regression is the best.Therefore,support vector machine regression is selected for camouflage color reconstruction.Finally,based on the results of camouflage spectral reflectance,it is found that the local sample set reconstruction based on self-training has higher color accuracy.The local training regression method based on LSSVM algorithm is used to realize the color reconstruction of five camouflage fabrics.Compared with the results of Datacolor,the average color difference is 4.16,3.24,2.65,3.54 and 2.36,respectively.Compared with the results of DigiEye,the average color difference is 3.06,3.58,3.71,3.77 and 3.34.The results show that the multispectral technique measurement method proposed in this paper is feasible for color evaluation of camouflage. |
PDF Full Text Request