Optimal Design Of X-ray Vision Detection System And Research On Step Sampling Mechanism

At present, Most of product testing with multiple parts in its internal adopt fixed ray imaging system and product continuous rotation to obtain information about the images of internal parts by identifying the fixed sampling step length of minimal size target. Whether the product is qualified is based on the human eye observation mode. Fixed-step capture images will get a lot of redundant information, which greatly reduces the detection speed. The human-eye observation mode is susceptible to subjective factors. The detection speed is slow and the test results can not guarantee high stability and high reliability. When building the inspection system, the factory has no reference system about the selection of the imaging device, which causes unnecessary waste.This paper based on the above background made the following research. First, we carried out a detailed study for the resolution of the system. The relationship between inherent unintelligibility and the smallest size of defect was derived when radiation source focal spot size was larger than the minimum size of defect, getting the best focus size and minimum system inherent resolution. They provided a reference to optimize the design of the X-ray stereoscopic vision detection system.Secondly, the paper regards the size of the inspecting objects, radian distortion and spatial distortion as the priori information, then function relation among them was derived from the theory. The correctness of the function was verified by experiments. Based on it, the sample size of each inspecting objects can be determined. Calculating the similarity for qualified and unqualified products, we can determine the optimum target detection range of inspecting objects and the optimal threshold. We relied on the positional relationship of inspecting objects to select the best detection method. Finally, the improved SIFT algorithm combines three search method were used to quickly find the products which were randomly placed in the image sequence standard library.Finally, we used the variable step sampling mechanism and detection strategy with combination of serial and parallel to do automatic detection for the inspecting objects of products which having four different structure sizes. Experiments proved that compared with the traditional algorithm, this algorithm has faster detection speed and robustness without losses of accuracy.