Space-borne high signal-to-noise ratio (SNR) and high spectral resolution spectral detection system with high detection accuracy (1-4ppm) is demonstrated under the application background of the detection of atmospheric carbon dioxide as the main component of greenhouse gases. According to greenhouse gas concentrations detection accuracy requirements and simulation of different spectral absorption spectrum of carbon dioxide, the reasonable spectral channel center wavelength, spectral bandwidth and spectral resolution is determined of the high spectral resolution carbon dioxide remote sensing system. Grating spectral imaging system using large area diffractive grating spectral as a core splitting element is to achieve fine spectrum splitting. By the application of large area array detector push-broom mode, the hyperspectral greenhouse gas detection system is developed with the spectrum center wavelength of 0.76um, 1.61um and 2.06um, spectral resolution indicators better than 0.047nm, 0.142nm and 0.182nm actually. The system components and working principle are described. Important parts involved in the system design such as spectral imaging system, large-array CCD visible-light detector, large-array HgCdTe infrared detectors, high SNR and low temperature drift imaging electronics, etc. are discussed. SNR indicators of three spectral ranges are estimated based on system parameters, in order to analyzing realizability of high detection accuracy of XCO2. The system performances are tested by taking fine spectral calibration and radiometric calibration methods in the laboratory. Spectral calibration results showed that: three spectral channels mean spectral resolutions of hyperspectral detection of greenhouse gases are better than 0.042 nm, 0.128nm and 0.17nm, three spectral channels average SNRs are up to 53dB, 48dB and 45dB respectively under the typical operating conditions of system. Development of this system successfully filled greenhouse gas detection systems onboard the blank in China.