This paper describe a research theoretically of the conversion result to the surface temperature based on long wave infrared detector, proposed a temperature measurement, then validate it by experiments. First, it introduces the constitution and measurement principle of the medical infrared thermal imager. Then, the conversion drift characteristic of infrared detect is described, the experimental data under variable environment is analyzed, and a temperature measurement and a drift compensation formula is proposed. Finally, some experiment with black body was accomplished. The results show the temperature error is under 0.3°C, confirm the validity of the measurement.
In order to solve the problem of the existing single waveband thermal imaging system can’t get precise temperature of object with emissivity unknown, an optical system of beam splitting lens and filter were used to established a colorimetric temperature measurement system based on infrared thermal imaging system. Completed the compensation for non-effective pixel, enhancement of contrast, calibration of nonhomogeneity and coherence for infrared thermal imaging system according to the application requirement, then acquired the calibration data with blackbody as radiation source at 200°~500° and fit it. A temperature measurement test performed at last, compared with the result acquired by thermocouple and single waveband thermal imaging system, it was shown that the colorimetric pyrometry system achieve the attractive precision after calibration and applied to measure the temperature of the object with emissivity unknown.
This paper proposed a new algorithm of inter-frame filtering in IR image based on threshold value for the purpose of solving image blur and smear brought by traditional inter-frame filtering algorithm. At first, it finds out causes of image blur and smear by analyzing general inter-frame filtering algorithm and dynamic inter-frame filtering algorithm, hence to bring up a new kind of time-domain filter. In order to obtain coefficients of the filter, it firstly gets difference image of present image and previous image, and then, it gets noisy threshold value by analyzing difference image with probability analysis method. The relationship between difference image and threshold value helps obtaining the coefficients of filter. At last, inter-frame filtering method is adopted to process pixels interrupted by noise. The experimental result shows that this algorithm has successfully repressed IR image blur and smear, and NETD tested by traditional inter filtering algorithm and the new algorithm are respectively 78mK and 70mK, which shows it has a better noise reduction performance than traditional ones. The algorithm is not only applied to still image, but also to sports image. As a new algorithm with great practical value, it is easy to achieve on FPGA, of excellent real-time performance and it effectively extends application scope of time domain filtering algorithm.
Traditional bad pixel detection algorithm is always based on the radiometric calibration. This method is easy to operate, but only suitable for the bad pixels whose positions are fixed. During the longtime operation period, environment temperature usually has drastic influence on IRFPA, the number of bad pixels often increase and their positions also vary, this result in the degradation of infrared image quality. In this paper, a new scene-based adaptive bad pixel detection algorithm is proposed for IRFPA. The algorithm firstly comparing the pixel value with its neighborhood, and affirm bad pixels preliminary through a suitable threshold. Then the potential bad pixels from different scene are matched, false bad pixels caused by scene and targets are eliminated, real bad pixels are confirmed. The essence of the proposed algorithm is using the correlation between the pixel and its neighborhood. The bad pixels and some targets in the scene have a weak correlation within neighborhoods, and the position of bad pixels varies slowly while the scene varies drastically when IRFPA is in use. This new method can be implemented in hardware easily and achieve the real time demand. With the real infrared images obtained from a camera, the experiment results show the effectiveness of the proposed algorithm.