Infrared Images In The Weak Multi-target Detection

Infrared imaging system was increasingly applied in the measurement of shooting range. Exactly extracting the offset of every weak small target in infrared image was an important part of modern range optical measurement equipment. In high speed image acquiring system, in order to accelerate to report shooting result, we need to find the useful sequence which contain targets from large amount images, and to process the useful sequence by right image segmentation method, then to interpretate the offset of every target. The key of this process was detection of small-weak signal. So the author focused research on the detection and recognition algorithms of weak-small-multi targets in infrared images.First, the definition and characteristics of weak-small-muti targets in infrared image were discussed. The author showed the model of infrared image and analysed the characteristics of noise, background and target. Some effective algorithm based on suppressed background, such as median filter, unsharp mask, top-hat transformation of mathematical morphology operations, were introduced for the target detection in single image and sequence images. It showed with the simulation results that the presented algorithm has excellent performance. A detection algorithm based on Contrast box was proposed. Image was segmented by double windows and contrast threshold. The method can efficiently segment the low SNR weak-small-multi targets in the infrared image of the clouds sky background, and suppress the noise. Simulation results demonstrated that the presented algorithm was good and simplely to operate. Detection probability was improved. Some image segmentation algorithm such as histogram method and based on Haralick cubic model to approximate locally image intensity method were also discussed. Simulation results showed that the Haralick model algorithm can meet the demand of task.A useful automatic interpretation method of weak-small-multi targets in infrared image based on sequence images was proposed. Structure of system and software flow were discussed. First, an automatic target sequence search method was proposed, the search result was analysed. Then the sequence image was segmented and interpreted using Contrast box algorithm. Finally, interpretation theory was introduced. Software flow and primary function were explained.It was proved by experiment that manual interpretation and half-autoatmic interpretation were necessary complementarity to automatic interpretation,the time and precision of interpretation were greatly improved.