Research On Moving Infrared Weak Small Target Detection Method

Moving infrared weak small target detection is the key link of infrared search and tracking system,which has extremely important application value in the fields of national defense and security,autonomous driving and other fields.Small targets have the characteristics of small image proportion and lack of geometric features.It is a difficult task to complete the detection of moving infrared weak small targets.At present,existing algorithms have high false alarm rates and high computational complexity.To address these issues,improvements have been proposed for single frame and multi frame algorithms.The main research contents are as follows:(i)To solve the problems of high computational complexity and high false alarm rate of multi-scale local energy factor,a detection algorithm based on double-layers energy factor(DLEF)was proposed.Firstly,the DLEF factor and WLF were used to process the original image.Next the two-dimensional Gaussian kernel was used to fuse the results of above.Finally,adaptive threshold segmentation was performed to complete the detection.Experiments on a single frame dataset showed that the detection algorithm based on DLEF has a certain improvement in object detection ability compared to mainstream single frame detection algorithms.In terms of real-time performance,the time consumption of multi-scale local energy detection algorithms was reduced to one-third.(ii)Aiming at the problems of low detection accuracy and high computational complexity in particle filter detection algorithm,a two-stage traceless particle filter algorithm(TUPF)was proposed.Firstly,in the importance sampling phase,the UKF based on the proportional symmetric sampling strategy was used to obtain the recommended distribution.In the phase of particle weight update,the quality factor Q was introduced to redistribute the sample weight,which effectively slowed down the particle degradation rate.Experiments on sequential datasets have shown that compared with the same series of algorithms,the TUPF algorithm has a certain improvement in the mean square error index.In terms of computational time,the TUPF algorithm reduces the resampling phase time by 10-20%.(iii)In actual environments,there are high requirements for weak small target detection algorithms in terms of algorithm complexity and robustness.Using only a single frame algorithm has fast computing speed,but relatively weak detection ability.Using only a multi frame algorithm consumes large computational resources,but has strong detection ability.Combining the above two,a single-multi-frame fusion infrared weak small target detection system was constructed.This system consisted of target detection module and trajectory tracking module.In the target detection module,phased detection was used to reduce the computational resource consumption of the all-weather system,and in the trajectory tracking module,insertion of a single frame algorithm was used to improve trajectory accuracy.Through simulation experiments on data collected by an infrared imaging equipment,the trajectory tracking effects of the fusion algorithm and a single algorithm were compared,the calculation time of each module was analyzed,and the detection effect of the system was tested,verifying the usability of the system.