To address the problem of continuous and accurate measurement of atmospheric NO₂ at night, a nighttime atmospheric NO₂ detection system based on target reflection light is developed. The system mainly consists of a transmitting unit and a signal receiving unit, with a 3.5 W high-power semiconductor laser with a central wavelength of 445 nm as its light source, and according to the laser differential absorption characteristics of NO₂ in the range of 440-450 nm, the laser detection wavelengths for NO₂ are determined to be $\lambda$_on = 444.8 nm and $\lambda$_off = 446.7 nm. Firstly, the sensitivity of wavelength and intensity of the semiconductor laser to its temperature and current was studied, and the quantitative relationship between the wavelength and intensity of laser and the temperature and current of laser was determined. Then basd on the quantitative relationship, the laser wavelength can be adjusted by controlling the temperature stability of the semiconductor laser while changing the current, and the influence of laser wavelength adjustment on laser intensity changes during the detection process can be eliminated. Finally, based on the constructed system, NO₂ sample gas experiments were conducted, and the fitting values of NO₂ differential absorption cross section were determined. After the determination of system parameters, field experiments were carried out at Anhui University and Science Island (the campus of Hefei Institutes of Physical Science, Chinese Academy of Sciences) respectively. NO₂ nighttime concentration near the ground at two observation points was successfully obtained, and the measurement results were compared with the data from National Environmental Monitoring Station and the data from the long-path differential optical absorption spectroscopy (LP-DOAS) at Science Island respectively. It was found that the observation data of the developed system showed good consistency with the data of National Environmental Monitoring Station and the data of LP-DOAS, with coefficients of determination of 0.902 and 0.891, respectively, verifying the accuracy and reliability of the system. Furthermore, NO₂ vertical distribution detection experiments were conducted, and NO₂ concentration values at different heights were successfully obtained, proving the feasibility of the developed system in NO₂ vertical distribution detection.