| Forest color is an important factor to measure forest landscape quality,and also a critical evaluation index in forest aesthetic evaluation.Against the strategic background of ecological civilization and“Beautiful China”,status of forest color landscape construction has been improved gradually.Therefore,it is significant to systematically and scientifically study technical level of improving forest color ornamental effects and color quantification,in order to contribute practical guidance for creating scientific forest color landscapes and improving seasonal landscape configuration and spatial pattern configuration of colors.This study mainly discussed single trees with autumn-colored leaves and forest landscape color in Hohhot,Fuyang in Hangzhou,and Chun’an,expecting to analyze key issues related to forest color quantification and forest aesthetic evaluation.Through field colourimetry and document collection,it explored and compared the characteristics of different colourimetry methods,and found forest color conversion methods with different color spaces and different colourimetry methods.At the same time,NCS(nature color space)was established so that conclusions of different researchers could be converted,compared and shared mutually.On the other hand,with process of forest color formation,transmission and reception as main research line,relationship between color and environmental factors in forest color formation was explored from the angle of color visual communication and forest landscape aesthetics,to reveal key environmental factors affecting color change of forest.Moreover,based on psychophysical method and public evaluation,this study,from perspective of forest aesthetics,evaluated beauty degree of relationship between forest color and change of viewing distance during transmission,and different ornamental effects of forest landscape with variation of color attributes,quantity,and spatial distribution during receiving process.In addition,it constructed and screened forest color indicators that affected aesthetic quality of forest landscapes under different color vision modes;explained the relationship between seasonal changes of forest,visual distance of forest colors,forest color landscape patterns with ornamental effects;displayed key influencing indicators of ornamental effects;established linearity and nonlinear models to refine forest color ornamental effect combinations of the optimal model.This is hoped to provide a reference for deepening forest color research and improving evaluation index system for color landscape ornamental effect.Conclusions were drawn as follows:(1)Conversion of different colourimetry methods for forest colors.According to study,measurements were significantly different under four colourimetry methods(photograph,scan,colorimeter,NCS colorimetric card).In order to mutually convert different measurement results,this study compared accuracy in converting forest color values by three color conversion methods-Photoshop formula method,multiple regression method,and the BP neural network(optimal BP model,R-ELM,SVM)and 5 models.Results showed optimal BP neural network obtained by self-compiled program had the highest conversion accuracy(R~2=1,MSE=0.0015～0.1481,running velocity up to 0.02s),which shall be particularly recommended to be applied at current stages.Set network parameters:trainrp function in BP network was operating function,with number of training times 1000,minimum error of the training target 0;network structure is 1 hidden layer including 10 neurons,1 input layer and 3 neurons,1 output layer and 3 output nodes and the learning rate 0.000001.(2)Establish forest color space.After field colourimetry and literature collection,the study summarized color information of 3,264 forest trees(leaves,flowers,fruits,branches)in the State.RGB and HSV color space values under different colourimetry methods of forest colors were converted into NCS color values by virtue of optimal BP neural network,realizing accurate conversion between NCS with RGB,HSV color spaces.On the basis of NCS color system,an attempt was made to construct national forest color space that served as subset of NCS space,including 879 forest colors obtained,which were distributed between NCS S 0515-G90Y—NCS S 8500-N in 12 color systems and 37 hues,accounting for 45%of the NCS.Also,it investigated color attributes,and summarized hue,color saturation,and value of forest colors in different seasons and regions,In the meantime,NCS was divided into 12:4:4 color intervals by using visual equidistance,and 192simplified spaces of colors were acquired,proving that visual isometric division of color space was able to more accurately express the perception of colors when the colors were quantified.(3)Construction and screening of forest color index.Under the guidance of correlation between color index and color ornamental effect,advantages and disadvantages of different color spaces(RGB,HSV)and screen color classification plans were compared and explained.According to findings,HSV was superior to RGB,and three-classification scheme(dominant color+assistant color+embellishment color)was slightly better than four-classification.The color difference index simplified three-element index of color into one-dimensional index,with clear physical meaning,and this would realize dimensionality reduction and quantification of color index,and improve evaluation efficiency.In terms of correlation between color difference formula and SBE,external color difference index was better than interior index,with biased index formula>Euclidean weighted color difference formula>Euclidean color difference formula>conical color difference formula.Five special effect indicators(|r|>0.500)that affected SBE referred to saturation of assistant color(r=0.539),saturation of whole color(r=0.533),biased exponent of dominant color and assistant color(r=-0.515),relative area of dominant color(r=-0.521,),relative area of whole color(r=-0.519).A total of 12 effective indicators(|r|>0.300)were obtained-saturation,chromatic aberration,interior chromatic aberration Euclidean,external chromatic aberration Euclidean,deviation and numbers.SBE evaluation model consisted of six key indicators such as chromatic aberration of dominant color and assistant color hue,saturation of whole color,biased exponent of dominant color and assistant color,the external chromatic aberration Euclidean of dominant color and assistant color,the deviation of whole value,and relative area of whole color.The results showed trees had a high ratio of hanging leaves,hues closer to yellow-green,a pure bright whole color,and distinct gradations of values.Among saturation of whole color and value of dominant color and assistant color,the latter was harmonic color sceneries,while the former had strongly contrasting saturation,so ornamental effect was good.(4)Seasonal effect of forest color.In this regard,this study investigated how color phenology of 38 common autumn-colored trees changed with environmental factors in Hohhot,obtaining the best viewing period and chromatogram of color seasonal change.According to residual autoregressive analysis,color change rate of leaves was greatly affected by cumulative accumulated temperature,followed by average temperature.In other words,the higher the cumulative accumulated temperature and lower the average temperature,the faster the color change.NCS color system was introduced to calibrate color system of leaves,and analyze color change law of hue,value and saturation in basic color period,early discoloration period,middle discoloration period,full discoloration period,and end discoloration period of trees with autumn-colored leaves.Based on evaluation of beauty degree of trees with autumn-colored leaves,it concluded ranking of ornamental effects,and explained differences in beauty degree in different periods.In fact,colors could be divided into ascending,descending,and Tablele colors in light of sequential changes of SBE value.Subsequently,cluster analysis was carried out in accordance with beauty degree and duration of color viewing period,with 38 types of trees classified into 6 levels:long-term single ornament,high-quality medium-term,combined medium-term,long-term grouping ornament,and auxiliary short-term.(5)Forest color perspective effect.Results told that when viewing distance increased as a whole,the color hue(H)gradually enhanced,while saturation(S)and value(V)declined slowly.The increase in viewing distance made whole scene become darker and blacker gradually,and saturation was more and more dimmed until it lost hue and became black;value tended to be black;the hue changed from yellow-green to deep blue-purple.With increase of viewing distance,the Scenic beauty estimation of forest colors decreases.When the viewing distance was<300m,the sceneries were basically clear,and morphological characteristics between trees could be distinguished.When it was300-700m,sceneries were relatively blurry and the sense of boundary between trees faded away.At viewing distance greater than 700m,it was difficult to identify single trees,and forest colors of the picture were in patches or groups.(6)Effect of forest color landscape pattern.There was a significant and most significant correlation between SBE and color pattern index.The color ornamental effect was significantly(P<0.05)or highly significantly negatively(P<0.01)correlated with relative area of the largest patch(r=-0.311),forest color distribution pattern(r=-0.690)and contagion index(r=-0.398).But it was significantly or highly significantly positively correlated with average patch fractional dimension(r=0.322),Shannon’s diversity index(r r=0.352),landscape abundance(r=0.655),and value of whole color(r=0.482).In accordance with regression analysis,ornamental effects were higher of forest color landscape with large relative area of whole color but small relative area of dominant color;naturally circuitous edge of color spot and shapes tending to be full;abundant colors,but uneven spatial distribution to gather but incompletely connected as one,and with color spots embedded naturally and orderly.This study proposes forest color distribution model index,which affects SBE significantly(r=-0.690).SBE is particularly high for color distribution models with orderly density changes in color pattern,obvious sense of order,and strong naturalness. |
