Research On Target Detection And Recognition Technology Based On Machine Learning

In the fields of computer vision,image processing,pattern recognition and machine learning research,target detection and target recognition have always been hot topics forever.Due to their unique advantages,infrared imaging and processing technology are widely used in intelligent transportation,security monitoring,industrial production detection,military target identification,etc.In the scientific research journey,we have been thinking about whether machines can have the thinking ability and intelligent problem solving like humans.With years of continuous research,machine learning algorithms have made great progress,especially for target detection and recognition in the field of infrared technology.The research work has achieved great success,but there are still many key scientific and technical problems that need to be solved.The environment in which the target is located is complex and changeable.At present,there is no relatively mature detection and recognition algorithm.Therefore,in practical applications,algorithm research is both opportunities and challenges.This article studies the three aspects of the infrared target detection and recognition technology application fields,the main innovative results of the research are as follows:1.Research on Infrared Small Target Detection and Recognition Based on Contrast:When the shape of the target itself is very small,or the distance between the target and the infrared imaging system is very far,the imaging area of the target is small,and it appears as a solitary point or spot in the field of view.It is difficult to detect and identify infrared small targets.This paper proposes MLCM algorithm based on local contrast measurement method,combined with SSDA template matching algorithm,mathematical morphology operation and MPCM algorithm to solve the problem of being a solitary point or spot target in the field of view.The algorithm detects and recognizes infrared point targets.The algorithm has good performance and is suitable for a variety of application scenarios.2.Research on Classification and Recognition of Infrared Area Target Based on Support Vector Machine: The area target has larger imaging area than small target.In the infrared imaging system,the same person wears different clothing,the imaging characteristics are different,and the characteristics of different pedestrian imaging are also different.This paper adopts a classification model based on SVM+HOG,extracts the target texture feature by adding the GLCM algorithm,classifies and recognizes infrared surface targets,and solves the problem of classification and recognition of infrared surface targets wearing different clothing.The results show that in the longwave infrared surface target scenes collected by myself,the correct rate of the proposed SVM+HOG+GLCM algorithm model for the classification and recognition of surface targets reaches 90.5%,which can meet certain application requirements..3.Research on Infrared Aerial Moving Target and Pedestrian Detection Based on Convolutional Neural Network: For the detection of infrared aerial moving targets and pedestrians,two approaches are used to solve them.For the infrared aerial moving targets,the traditional inter-frame difference detection algorithm based on threshold segmentation is used first;then the Faster R-CNN algorithm is used to detect it;finally,it compares the detection results of the two types of algorithms and analysis their advantages and disadvantages.For the pedestrians,it also uses the traditional cascade detector to detect the target;then it uses the Faster R-CNN algorithm to detect the target;finally,it compares the detection results of the two types of algorithms.According to the detection results of the two target scenes,the detection effect of the CNN algorithm is more excellent.In short,this article mainly studies the three aspects of target detection and recognition in the field of infrared technology,and it also studies the traditional target detection algorithm,then it compares the detection results of the two types of algorithms,and it studies the advantages and disadvantages of the algorithms,it obtaines certain research results,but there are still some problems,especially machine learning algorithms are a new type of research field that has emerged internationally in recent years,but it need to be further deepened in future research work.The research results in this article will play a certain role in promoting the application of target detection and recognition in the field of infrared technology.