By using image analysis technology and automatic control technology, the artificial intelligence technology was applied to the detection of unsound wheat kernels. The automatic, rapid and nondestructive detection instrument for unsound wheat kernels was developed. Through the verification of accuracy, repeatability, stability, difference between two instruments and other related performance parameters, the results showed that the instrument had stable performance. The accuracy, repeatability, stability and difference between two instruments met the requirements of industry standards. The operation was simple, and the detection speed was fast. It can overcome the problems of strong subjectivity, poor repeatability and poor inspection consistency among different personnel with manual checking, achieving the automatic and rapid nondestructive testing of unsound wheat kernels, and meeting the testing demand of grain collection and storage enterprises, processing enterprises and inspection agencies.