Abstract: Abstract: The optimized operational parameters of a combine harvester are often set once at the beginning of the operation which makes it very inconvenient to observe the grain purity and adjust these parameters frequently during the harvesting process. It is imperative to develop an instrument which can obtain grain impurity data for optimizing operational parameters during the harvesting process automatically. Based the purpose above, a grain impurity (GI) sensor was designed. The sensor consisted of electromagnet for the extension control of choke panel, spring for the reset of choke panel, visual glass window, choke panel, light resource and camera which is used for image capture of grains gathered by choke panel. Then the captured image was processed for calculating the grain impurity. The installation position and illumination type of LED light source was optimized to obtain high-quality images according to the image histogram. The pulling force applied to the choke panel by the electromagnet, the thrust force applied to the choke panel by the spring and the entrance size of sampling box were analyzed. The results showed that illuminating the visual window indirectly with two LEDs installed on both sides of the sampling box was a better way, with the peak in the image intensity histogram avoided effectively. The electromagnet with a rated pull force of 60 N could provide enough pulling force and the spring with diameter of 1 mm could provide enough thrust force. Due to the fact that the short stalks and straw during harvesting ranged from 10 to 30 mm in length, the sampling box with height of 95.7 mm and width of 76.5 mm was designed. And the inlet width was 31.9 mm and the visual window' length and width wan 57.1 and 57.4 mm respectively. Therefore, the grain, stalk and straw could flow into the sampling box smoothly and about 200 grain kernels could be seen in the visual window. The monitoring accuracy of the sensor was verified with different impurity contents in. bench test and the results showed that the device was capable of monitoring the grain impurity content in the range of 0-2.88%, whose change trend was consistent with the artificial grain impurity content. A dust-proof cover was designed to provide a stable working environment for the camera in the field. And the field experiments were conducted with YAMMA 4LZ2.5 combine harvester in Jiuli Lake in city of Suzhou, on Nov.17, 2017. The results showed that the relative errors of grain impurity in field test were between 9.44% and 19.67%.