Research And Application Of Infrared Object Detection Algorithm Based On Learning

With the development of computer vision technology,object detection,as one of the basic topics of computer vision,has received more and more attention from researchers.In real life,due to the influence of background radiation,interference objects and other factors,the object features in the infrared image change greatly,which brings great challenges to object detection.Aiming at the main problems of infrared object detection of leaking gas and the key point of fixed-wing aircraft,this paper studies the infrared object detection method based on learning.Aiming at the problem of low detection accuracy of leakage gas due to interference from environment and moving objects in chemical production,a method based on learning infrared leak gas target detection is studied.In this paper,the suspected leakage area in the leaked video is extracted by the appropriate background modeling method,and the denoising is performed by the connected domain analysis.Then,the shape characteristics of the leaking gas are selected for extraction.The model is trained by the support vector machine classifier(SVC)in machine learning.The model is used for classification during the detection,so as to detect the leaking gas.Finally,a comparative experiment is carried out.In the security work,due to the high speed of the aircraft and the change of attitude,the detection accuracy of the key point of fixed-wing aircraft in the infrared image is low and the speed is slow.The learning-based detection method of the key point detection and tracking of fixed-wing aircraft is studied.This paper proposes improved detection and tracking algorithms based on SVC/SVR.In the detection and tracking method based on SVC+KCF,the detection module first extracts the fixed-wing aircraft region by image segmentation,and then selects the fixed-wing aircraft features for extraction.Then,the SVC is used to judge the flight direction of the fixed-wing aircraft,and then the key point positioning is performed according to the orientation,and the tracking module uses the KCF algorithm.Based on the SVR+KCF detection and tracking method,the detection module uses the SVR to directly obtain the coordinates of the key points from the features of the fixed-wing aircraft,and the tracking module does not change.Finally,a comparative experiment is conducted.The three infrared videos are used to test the learning-based leaking gas detection method.The results show that the method can achieve a detection rate of up to 98%,which is better than the texture based and Adaboost method.It can be seen that the leak gas detection method has advantages in chemical leak gas detection.Testing the key point detection method of a fixed-wing aircraft based on learning using a 3 fixed-wing aircraft infrared videos.The results show that the SVC+KCF method has an accuracy of86.18%,an error mean of 21.70,and a frame rate of 39.78 frames/sec.In the case of a certain accuracy,the speed is faster,and the SVR+KCF method has an accuracy of96.77% and an error mean of 7.93.The frame rate is 25.34 frames/sec,which guarantees a certain real-time performance and high accuracy.The method of this paper has an advantage in the detection of the key point of fixed-wing aircraft in infrared images.