Fine spectral information of hyper-spectral remote sensing provides a good prospect for parameters of water color inversion by remote sensing. However, for high spectral resolution and low spatial resolution, the current hyper-spectral remote sensing inversion models and algorithms of water color are generally lack of effective use of spatial information, so it is difficult to ensure accuracy and stability of the model. Taking Chaohu Lake, Anhui, China, as the study area, based on the space eight neighborhoods and genetic algorithm, hyper-spectral remote sensing HSI data of HJ-1A satellite is combined with the measured sample data to construct a hyper-spectral remote sensing inversion model of chlorophyll by the in-depth analysis of the spectral characteristics of water body. Based on matlab7.0 platform, the parameters of inversion model is calculated by the combined spectral indices and genetic algorithm. Under the spatial neighborhood analysis and genetic iteration, the concentration of chlorophyll is solved. The results show that the genetic algorithm abandons the traditional search methods, optimizes and searches water color space randomly using simulated evolution in the vicinity of the spatial domain by spectral information, jumps out of the local extreme point, can effectively improve the accuracy of model inversion.