Summary of application of hog features

Directional gradient histogram (histogram of oriented Gradient, HOG) features advantages and disadvantages Description Advantage First of all, since the HOG is operated on the local cell of the image, it can maintain a good invariance to the geometric and optical deformation of the image. These two types of deformation will only appear in the larger space area. Secondly, in the rough airspace sampling, fine direction sampling and strong local optical normalization conditions, as long as the pedestrian generally can maintain upright posture, can allow pedestrians have some subtle limb movements, these subtle actions can be ignored without affecting the detection effect. Therefore, the Hog feature is particularly suitable for human detection in images.
Pedestrian detection: Because the Hog feature can reflect the contours of the human body, and it is not sensitive to the brightness and color changes of the human body in the image, it has excellent performance in its detection. Vehicle detection: Due to the hog characteristics of the light is not sensitive, even if there is partial occlusion can be detected, in a variety of complex traffic road and parking lot, has a good robustness. Tracking aspect: The Hog feature in tracking the moving target with obvious edge contour hog represents the edge (gradient) structure characteristic, so it can describe the local shape information, and the quantization of position and direction space can restrain the influence of translation and rotation in some degree; take the normalized histogram in local area, Can partially counteract the effects of light changes. Due to some neglect of the effect of illumination color on the image, the dimensionality of the representation data required by the image is reduced. And because of this kind of sub-unit processing method, it also makes the relationship between the local pixel points of the image can be well obtained.
The shortcomings of the description sub-generation process, resulting in slow speed, poor real-time, difficult to deal with occlusion problems. Due to the nature of the gradient, the descriptors are quite sensitive to noise.
Pedestrian detection for pedestrian detection, vehicle detection, tracking, still image, video