Accurately obtaining the spatial distribution data of seeds after sowing can provide plant position information for field management, thereby improving resource utilization. Machine vision technology was applied to the detection of sowing position. After the seeds fell into the seed bed through the opener seed guide tube, the industrial camera was used to collect the image of the seed position in the seed bed, and the target detection algorithm was used to identify the seed position. However, during the operation of the traditional opener, after the seeds fell into the seed bed, the soil fell back to the seed bed too fast, which made it difficult to collect the original images of the seeds in the seed bed before the soil cover. Aiming at this problem, a soil backfill delayed opening device was proposed and designed. The design extended the soil backfilling time through the cooperation of the soil vectoring device, the seed pressing device and the soil backfilling device, forming an avoiding space for the original image acquisition. After the image acquisition was completed, the soil would be pushed back to the seed bed to ensure the soil backfilling rate and achieve the purpose of delaying soil backfill. The structural parameters of key devices such as soil vectoring device and soil backfilling device were determined through theoretical analysis. A discrete-element simulation test of soil backfill rate was carried out. Taking working speed, operating depth and rotation angle of soil backfilling plate as test factors, the discrete element simulation quadratic rotation orthogonal combination test of soil backfilling rate was carried out. The optimal combination of operation parameters was determined as the working speed was 1.6m/s, the operating depth was 30mm and the rotation angle of soil backfilling plate was 40°. The field test of soil backfill rate and seed image collection were carried out. Under the optimal parameter combination, the soil backfill rate of the soil backfill delayed opening device was 96.5%, which was 39.6 percentage points higher than that of the opener without the soil backfilling device. In the avoidance space formed by the designed soil vectoring device, the original image of the seeds in the seedbed can be collected by using an industrial camera. The experiment results showed that the structure designed can effectively avoid the impact of falling soil on image collection, ensure the soil backfill rate, and achieve the collection of seed images in the seed bed, which laid a foundation for computer vision technology to detect the quality of seeding operations.