Classification from remote sensing imagery has been proved to be a quick and effective method for automatic mapping of vegetation population.However，the interference from background noises and the weak spectral separability between vegetation populations have a negative impact on the identification of urban vegetation populations from remote sensing images.This make the classification accuracy unsatisfied as using conventional classification methods.In order to solve this problem，a new method of classification from remote sensing imagery，named as SLPA，is proposed.SLPA consists of two parts：Spectrum\|Location joint analysis (abbreviated as S\|L analysis or SL) and Phenology remote sensing Analysis (PA).By adding density descriptors to a feature space and by combining the descriptors derived from winter and summer images of a scene into the same feature space，these two kinds of analyses can be embedded into the classification process.The embedding increases the number of available independent descriptors therefore making the feature space rich enough to adapt to such a complex classification；meanwhile，uncertainty in the classification can be reduced and the classification accuracy can be improved significantly.The data from accuracy assessments show that with the S\|L analysis and the phenology analysis，the overall accuracies of the classification for urban vegetation population will increase by 15% and 29.3% respectively.In addition，the S\|L analysis almost does not increase the computational complexity because a frequently used computation for the density of gray elements is replaced with a calculation of binary mean value，a much more time saving operation.Experiments indicate that SLPA is good at robustness and universality in the classification of urban vegetation population.