Research On Infrared Image Target Detection Method

Due to its advantages of all-weather work,infrared image target detection technology has a very wide range of applications in intelligent driving,intelligent monitoring and other fields.Although infrared image target detection technology has made great progress,due to the diversity of pedestrian targets and the influence of complex scenes,the existing infrared image pedestrian detection methods still have poor results in extracting regions of interest(ROIs),a single feature cannot accurately describe the target and other problems.Therefore,the infrared image pedestrian target detection technology still has a large room for improvement,which has important research significance.Aiming at these existing problems,the main research work of this paper is as follows:(1)Aiming at the problems of poor threshold segmentation effect,over-segmentation,and a large number of redundant candidate boxes generated by target positioning in the region of interest method,a two-dimensional Otsu threshold segmentation based on the firefly algorithm and a connected region based on geometric parameter constraints are proposed.Marker-combined region-of-interest extraction method.The method firstly uses two-dimensional Otsu threshold segmentation to segment the image.When searching for the optimal threshold,the firefly algorithm with simulated annealing factor is used to optimize it to obtain the most accurate threshold.The segmented images are then processed morphologically.Finally,the processed image is marked by the connected region labeling method based on geometric parameter constraints,and the region of interest is obtained,that is,the region containing the pedestrian target.(2)Aiming at the problem that a single feature cannot accurately describe the target and the detection rate is low in different scenarios,an infrared image target detection method based on multi-feature fusion is proposed.Firstly,aiming at the problem of texture loss in infrared image target detection,LBP feature based on distance weighting is used,and distance weighting is performed by different distances between each pixel in the neighborhood and the central pixel.Aiming at the problem of insufficient edge extraction capability in infrared target detection,a local amplification HOG feature based on entropy weighting is used.First,the Bin channel of the HOG feature in the Cell is locally amplified,and then entropy weighting is performed on a Block as a unit.Finally,the weighted fusion of the two improved features is sent to the trained SVM classifier for classification,and the final detection result is obtained.Through the simulation comparison experiment and the subjective and objective analysis,it is verified that the method in this paper is advanced,accurate and robust.For the region of interest extraction task,the ROIs extraction method proposed in this paper,compared with other methods,has better segmentation effect,more accurate target positioning,and reduces the number of ROIs and improves the segmentation efficiency.For the target detection task,the infrared image target detection method based on multi-feature fusion proposed in this paper has better detection performance than other methods,and the TPR value in different scenarios reaches 96%.