Reserch On The Key Techniques Of Uncooled Infrared Thermal Imaging System

Infrared imaging system has enjoyed a boom in both military and civilian fields in recent years. With low price, high reliability, small size and low power consumption, uncooled infrared imaging system plays an important role in the development of infrared thermal imaging system. Based on basic theories related to uncooled infrared imaging system and researches on infrared image processing algorithm, an uncooled infrared thermal imaging system with real-time image processing functions was designed and developed in the paper, with good results.The non-uniformity of infrared image has a significant impact on image quality and visual effect. By digging deep into the causes for the development of heterogeneity, the paper analyzed the strengths and weaknesses of the correction-based NUC and scene-based NUC algorithm designed for IRFPA. On this basis, a kind of combined NUC method was presented. At first, based on temperature of the focal plane substrate at the powering- up moment, the stored gain and offset correction parameters of corresponding temperature intervals were extracted from the storage devices to preliminarily eliminate the non-uniformity of the detector. By analyzing the characteristics of image non-uniformity noise left after preliminary correction, the paper came up with the idea that the edge-preserving P-M filtering was used instead of the four-neighborhood mean filtering involved in conventional neural network algorithm to obtain expected image and reduce image edge error. Experiments showed this algorithm featured fast convergence speed and high correction accuracy. It worked effectively in avoiding the image degradation caused by response characteristic drifting of the infrared focal plane.Considering the low contrast and poor temperature resolution of infrared image, a local contrast enhancement algorithm for infrared image, based on bilateral filter, was presented. For the algorithm, hierarchical processing was first conducted on images with bilateral filter, thus generating background images with relatively large dynamic range and the detail layer images with lots of noises and detailed information. After dynamic range compression and local contrast stretching of the background images, low bit-wide images suitable for human eye recognition were generated. Then the detail layer images were processed with Bayesian threshold denoising. And meanwhile, the weak edge information was enhanced. Finally, the background layer images and detail layer images were processed with weighted operation, and the enhanced images were produced. Experiments were conducted on large numbers of images. Besides, qualitative and quantitative analysis was made of the enhanced effect, and comparison was made between this algorithm involved here and existing commonly-used enhancement algorithms. Experiments showed that this algorithm worked effectively in improving the local contrast of images and highlighting texture details. What’s more, it inhibited image noise and improved the overall visual effect.Based on the structural features and the output signal characteristics of the infrared detectors UL02152, a low-noise infrared imaging system was designed. FPGA chip was used to offer timing-drive signal for the detector. With the application of low-noise bias circuit and high-precision digital signal processing technologies, the system noises were effectively inhibited. Also the integrity and stability of output signals were secured. The removal of image blind spots, non-uniformity correction and image enhancement were completed in FPGA. The image delay was less than 2 frames, meeting the requirements on real-time output. Tests verified that the infrared image NETD was 74mk, which was up to the standards on low system noise.