In order to solve the problem that the accuracy of heart rate measurement is poor under external interference or dim environment by image Photoplethysmography (iPPG), an adaptive heart rate extraction algorithm was presented and was verified on an embedded hardware platform. Different scenes could be identified by this algorithm according to the chromaticity relationship between faces and the background areas in the image. Subsequently, the appropriate camera was launched for image acquisition and adaptive mapping. Finally, the extracted signal was filtered and the result was output after the signal quality evaluation. The algorithm was implemented on the embedded Zynq system with dual cameras for real-time heat rate measurement, and the results were visualized. Experimental results show that the measurement error decreases from 3.36BPM to 2.78BPM under the interference of light and motion, and the accuracy is improved by 17.3%. In addition, the designed system can also achieve heart rate acquisition under extreme dark conditions, with an average error of about 2.39BPM.