Open AccessImproved HOG Feature Vehicle Recognition Algorithm Based On Sliding Window
Author(s) -
Ji Yang,
Jun Zhong
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1627/1/012013
Subject(s) - sliding window protocol , histogram , computer science , artificial intelligence , histogram of oriented gradients , pattern recognition (psychology) , classifier (uml) , randomness , feature (linguistics) , algorithm , window (computing) , computer vision , mathematics , image (mathematics) , linguistics , philosophy , operating system , statistics
With the continuous development in recent years, machine learning has been extensively used in many application scenarios, especially in vehicle recognition. At present, the histogram of oriented gradients (HOG) algorithm has been widely used in vehicle recognition and facial recognition. This paper studies the HOG algorithm, introduces the vehicle recognition using the HOG algorithm, and introduces a detection method based on sliding window. In this research, an improved HOG algorithm based on sliding window is proposed for vehicle recognition and compared with the traditional HOG algorithm. In randomness test, the traditional recognition method often causes its HOG features to be trapped in a large number of invalid features due to the uncertainty of the vehicle distribution. This makes it difficult to be identified by the classifier, resulting in erroneous images, with an accuracy of only 83.33%. The optimized sliding window algorithm eliminates the influence of overlapping windows, and the number of accurately-recognized images is largely increased, with an accuracy of 91.02%.