Abstract: Seeding amount determines the distribution density of plants in the field, thereby posing a significant impact on the final yield of crops. Real-time detection of sowing amount is an inevitable trend in precise agriculture. However, the current technology is difficult to rapidly identify overlapping seeds for the total amount of wheat precision seeding. In this study, an accurate detection system with an interpolation sensor was developed to control the seeding amount in a wheat precision planter with a wide seedling strip. Infrared radiation LEDs were used as the sensitive elements, and the 24-bit ADS1256 as the high-precision sampling module. The sampling frequency of the system was determined to be 1×103 Hz. The process of seed falling was analyzed, where the seeds were sampled at least three times when the sensor was scanning over the detection areas. The sampling voltage of one seed, two seeds, and three seeds were obtained when the sensor passed through. Two sampling threshold divisions were set using the peak and the mean value, according to the sampling voltage. In the peak, 489 was the threshold value for 1 and 2 seeds while 616 for 2 and 3 seeds. In the mean, 277 was the threshold value for 1 and 2 seeds while 330 for 2 and 3 seeds. A detection system of precision seeding was realized to integrate with the interaction interface of touch screen personal-computer and drive control of a stepper motor. A field experiment of planter under static state was carried out to verify the detection accuracy of the system, the adaptability for different varieties of wheat and seeding frequency, and further to evaluate the two-threshold division. Three types of wheat were selected as the research objects, including Heng-Guan 35, Ji-Mai 22, and Cun-Mai 11. It was found that the peak detection was optimal, due to the low detection error and good adaptability to different wheat varieties. The peak detection was further revised, where the correction coefficient was 7.08%. A field experiment was also carried out to further verify the performance of the detection system under dynamic conditions. It was found that the absolute detection error of the system ranged from 1.12%-5.63%, and the mean absolute error was 3.12%, indicating a high detection accuracy for the high requirements under various sowing rates and operation speeds, as well as a high resistance to sunlight and dust interference. The finding can provide potential technical support for the design of seed tube sensor, and the rapid detection of seeding amount in the process of precision wheat sowing.