Sample representativeness is the premise to ensure the authenticity and accuracy of test results. The application scenario of unmanned detection is taken as the research object. Through the research on the uniformity of major heavy metals content detection in the sample processing of unmanned sample sampler, it is clear that mixing operation and sampling control are the necessary steps to ensure the representativeness of samples. By the method of combining dyeing simulation and actual sample verification, the key parameters such as the blending operation mode and parameters, the amount of samples crushed and tested for automatic sampling were systematically studied, and an optimal blending sampling scheme was obtained to ensure the representativeness of samples for unmanned heavy metal testing. The results showed that, the anchor type mountain paddle was used to stir 100 laps in revolution mode or repeat mix and separate samples for 3 times, and the minimum sample sampling amount for crushing was 150 g. The maximum crushed particle size of the sample was 1 mm and the minimum weight was 0.5g. This scheme was suitable for the representative detection of lead and cadmium elements in rice and wheat samples less than 6 kg. The research results provide technical supports for ensuring the representativeness and accuracy of heavy metal detection results in grain.