Moving Vehicle Detection Using Haar-like, LBP and a Machine Learning Adaboost Algorithm

Vehicular networking and computer vision-based distance estimation for VANET application using Raspberry Pi 3

“…There exist a fast and trainable method for object recognition, The Haar Cascade Classifier [10] this method also used by [11] with the detector trained by BIT-Dataset and conclude that Haar Cascade Classifier is a good candidate for object detector. Research by [12] combines the Haar detector with the K-nearest neighbour technique to identify the license plate of vehicles that claimed to offer great efficiency for practical use.…”

“…Haar Cascade adopted by [10] was a machine learning mechanism. The classifier evaluates the simple feature of an incoming image rather than processing the image by pixel.…”

“…Two-rectangle feature (a) [10] used to extract edge feature on image and three-rectangle feature (b) used to extract line feature on image. For example, Figure 3 shows an applying Haar basic feature shapes over an image of a vehicle.…”

“…The pixels sum of the black and white are summed up (by subtraction) and the difference is the value of the image feature underlying the mask Haar feature. To immediately evaluate the features and the author of [10] presented auxiliary Integral Image. Any large number of features extracted, their values can be quickly calculated with only four points of reference.…”

“…The detection comes from the cascaded classifier was fluctuating in terms of vehicle position on the image at each measurement frame every half second. Since the nature of the cascade classifier applied extensive numbers of Haar feature by sizes and shapes [10] in sliding window, the results produced resembles a random process which generates a stationary white noise added to the measurement, this part intended to investigate the characteristic of the noise whether it is white or not and to see if optimum Kalman filter can be exploited. White noise defined as a series of uncorrelated random variables following a continues distribution with zero mean and unity variance also a process is a white noise if its close-spaced time series sample is uncorrelated [18], moreover, if those samples were independent, it follows Gaussian distribution hence guaranteed for randomness.…”