For the detection of heavy metals in soil, conventional methods of chemical analysis are of many deficiencies, such as long time consuming, complex pretreatment process, and the strong acid used will cause secondary pollution to environment. However，energy dispersive X-Ray fluorescence spectrometry have many advantages, such as non-destruction, fast test, simple pretreatment, easy to carry etc., which make it especially suitable on rapid on-site monitoring. When using this method to detect cadmium, the energy of K series characteristic line is larger, making the limit of the detection relatively high for general EDXRF. Based on the energy dispersive X-Ray fluorescence spectroscopy, the rapid detection of trace cadmium in soil was studied. Appropriate devices were chosen to constructing the instrument platform, and the instrument structures and test conditions were optimized considering peak intensity and relative intensity of determined element.Research showed that the peak intensity increased linearly with the increase of tube current, while relative intensity had no obvious change. Hence, for the detection of cadmium, tube current chosen were as large as possible with the permission of X-Ray tube. Afterwards, considering peak intensity and relative intensity of cadmium with different thickness of filter and tube voltage, the optimized conditions could be ascertained by using theoretical relative deviation. And the optimal conditions were as follows: the tube voltage was 55 kV, tube current was 48 μA, the filter was molybdenum with the thickness of 1.25 mm. The measurement time had an influence on relative standard deviation of results. When measuring times were less than 500 seconds, the relative standard deviation of peak intensity decreased with time increasing, while having no more obvious change after measuring times longer than 500 seconds. And the relative standard deviation was smaller, the short-term precision of results better, which made repetition appropriate for the detection. So 500 seconds was chosen as measuring time. At the same time, test conditions of samples also had an influence on the results. Since both the peak intensity and relative intensity decreased with increasing thickness of film, the thickness of 12.5 μm of polyester film was chosen to be used. And both peak intensity and relative intensity of cadmium increased with increasing weight of samples at first, but when the weight was more than 3g, as the weight increased, the peak intensity of cadmium changed slowly and relative intensity had no obvious change. So the weight of samples should be more than 3 g. The peak intensity and relative intensity decreased as water content increased in soil, which meant that soil moisture would affect test results, and soil samples should be air drying or oven drying. Adopting the optimized conditions above, the linear correlation coefficient was 0.993 of samples when using national standard samples or samples determined by inductively coupled plasma mass spectrometry. A national standard sample was tested 11 times in which the concentration was 1.12 mg·kg-1, and the standard deviation of the results was 0.09, and the relative standard deviation was 8.22%. The limit of detection of this method was 0.16 mg·kg-1 detected by high purity silicon dioxide for 11 times, which was below the limit of the first grade soil. Compared with inductively coupled plasma mass spectrometry, the test results had significant consistency. After optimizing instrument structure and test conditions of samples, the limit of detection was lower for trace cadmium in soil based on energy dispersive fluorescence spectrometry, which will play an important part in the rapidly screening of pollution area and the measurement in large area of cadmium.