Tracking Objects Based on Multiple Particle Filters for Multipart Combined Moving Directions Information

Algorithm 6

Update gentle Adaboost.

	Input: Old strong classification F is a matrix, new D data set
	Output: A new strong classification
	Step 1:
	Initialize the weight set consisting of N equal numbers, that equal 1/N, where N is the number of elements in the data set.
	Initialize matrix size to contain weak classifications
	Initializes the chosen stack to save the position of the weak classification that have been selected
	Step 2:
	for i = 1 to s-T do
	Beginfor
	Initialize the loss stack to store error values
	for j = 1 to s do
	Beginfor
	if − then
	Calculate the error of classification j according to the following formula:
	Put into loss, loss.push()
	Endif
	Endfor
	Select the classification with the smallest error value,
	Put into loss, loss.push ()
	for j = 1 to s do
	Beginfor
	Update weight
	Endfor
	Endfor
	Step 3: freshly train weak classification T on data set by Algorithm 5
	Step 4: combine s-T weak classification in Step 2 and T weak classification in Step 3 to create a new strong classification .