It has been a long-standing and challenging goal to precisely measure water-leaving radiance (Lw) in ocean color remote sensing. Conventional approaches like in-water profile method, above-water method and other water measurement methods cannot directly measure Lw. Thus, they demand complex post-measurement processes, which cause many uncontrollable factors. Skylight-blocked approach (SBA), proposed by Zhongping Lee in 2013, was an innovative method to directly measure Lw, which can avoid uncertainties caused by post-measurement processes. However, no water spectral measurement system based on SBA has been developed so far. It is of great theoretical and practical significance to develop and test such equipment. Based on SBA, a water spectral measurement system was developed in this study. Firstly, the principles of the field water-spectrum measurement method, SBA is introduced. Then, the hardware structure and system design of the system are described in detail. The field experiment in Pearl River Estuary (113°32′38″E, 22°25′43″N) was carried out on September 20, 2017 to test the continuous measurement ability of the system. The system can measure Lw and downward irradiance of water (Es) directly and synchronously and then calculated the Rrs. The coefficients of variation (CV) of them are less than 5%, which basically proves the effectiveness of the water spectral measurement system in Lw measurement. The measurements results of the system are in good agreement with those of Maya2000 Pro synchronous. The continuous observation experiments show the stability of measurements and the ability to track the change of optical characteristics of water. The existing problems and future development prospects are pointed out, such as self-shading correction, data quality control, high frequency measurement, buoy tilts recording, multi-factor joint measurement, long time series and large range networking, etc. To summarize, the water spectral measurement system based on SBA called the Floating Optical Buoy (FOBY) can measure Lw directly in high the frequency of minutes which can track the rapid dynamic change of water optical characteristics. It is expected to improve the matching efficiency between in-situ measurement data and satellite remote sensing data. Based on the system, big data sets of water spectral can be obtained by means of the network. It is conducive to greatly improve the application potential of various satellite data on ocean color remote sensing.