Research On Infrared Image Enhancement Algorithms For Hot Forgings

Large forging manufacturing industry is related to national security and national economic lifeline,which is indispensable strategic industry and also an important symbol of a country's comprehensive national strength.The production characteristics of large forgings are huge shape and heavy weight,bad production environment,long production cycle and unit and small-batch production.So it is easy to cause the forging products is unqualified or low yield,and bring the huge waste of material and energy.In the process of forging,temperature field of the forgings is the main basis of deciding process parameters such as forging force,forging reduction and forging times.Accurately monitoring in the change of temperature field of the forgings and timely adjusting the forging process,which has an extremely significant significance for improving the yield and the utilization rate of material and energy.For large hot forgings,in order to get the temperature field of the object and have a higher measurement precision of temperature field,infrared image technique was used firstly and then post-processing was applied to the obtained infrared image.Because the hot forging is heavily influenced by bad environment factors in the production process,acquisition of images often appear all sorts of noise,distortion and blur,which results in severe decrease of accuracy of image detection analysis.To break the limit of inherent resolution for imaging system,from the perspective of software,some algorithms were proposed to improve the quality of infrared image of hot forgings in this paper.From two aspects,enhancement technologies of infrared color-image and infrared gray-image,which are applied to the infrared image processing of hot forgings.For the enhancement technology of infrared color-image,a single image super resolution technique based on Non Sub-Sampled Contourlet Transform(NSCT)and soft edge smoothness prior was proposed,which is based on the fact that every coefficient at the coarser level of the NSCT can be correlated with the coefficient at the next finer level in the similar orientation.Lorentzian error norm was used to reject the probable outliers in learning process,moreover,soft edge smoothness prior as a regularizing parameter,wasadopted to preserve the smoothness of smooth edges.The results confirm that this technique has better resolution with minimum complexity.For the enhancement technology of infrared gray-image,a new image resolution enhancement technique based on wavelet transform and bicubic interpolation was presented,which used discrete wavelet to decompose the original image into different subbands,and then the original image and these subbands have been bicubic interpolated.Meanwhile,an intermediate stage for estimating the high-frequency subbands was proposed in image processing,the difference between the LL subband image and the low-resolution input image was extracted and this difference image(High-frequency components)was used in the intermediate process to correct the estimated high-frequency components such as LH,HL and HH.At last,all these images were combined by using inverse wavelet transform.This method not only could effectively eliminate image noise and improve the image detail information,but also could preserve more edge information and obtain a sharper image.