To meet the miniaturization and integrated application of quantum random number generator (QRNG), a QRNG based on light emitting diode (LED) light source and complementary metal oxide semiconductor (CMOS) detection is designed. Quantum random numbers ensure the high security of random numbers based on the inherent characteristics of optical quantum, and their low cost and high speed processing methods also have high application value. However, traditional data sampling methods based on high-speed analog-to-digital converter (ADC) have the problems of high cost, complex system, and high post-processing requirements. In addition, the imperfection of ADC sampling also introduces pseudo-random characteristics, reducing the random number characteristics of the system, thus limiting its further commercial application and promotion. In the QRNG proposed in this paper, a binary method based on voltage comparison is proposed, which utilizes the randomness of the detection voltage between pixels, and uses the voltage difference between adjacent pulses in the output voltage sequence detected with CMOS pixels as a random number entropy source to obtain a quantum random number sequence. This method has the advantages of simplicity, reliability and easy realization, which is more conducive to the industrial promotion of products.