Use C++11 random generator instead of Boost in RangeLikelihood

[SunBlack]

Currently there is some Boost code, because pdf/cdf has no direct equivalent in C++14.

pcl/simulation/src/range_likelihood.cpp

Lines 559 to 584 in 93f9e72

	using boost::math::normal; // typedef provides default type is double.
	normal unit_norm_dist(0,1); // (default mean = zero, and standard deviation = unity)
	double costFunction5(double measured_depth, double model_disp, double sigma, double floor_proportion)
	{
	// NEED TO CONVERT MEASURED TO DISPARITY
	double measured_disp = (-0.7253/measured_depth + 1.0360 );
	// measured_depth = ref_val [m]
	// model_disp = depth_val [0-1]
	// upper and lower bound on depth buffer:
	double lower_bound =0;
	double upper_bound =1;
	double gaussian_part = pdf(unit_norm_dist, (measured_disp-model_disp)/sigma)/sigma;
	double truncation = 1/cdf(unit_norm_dist,(upper_bound-model_disp)/sigma) - cdf(unit_norm_dist, (lower_bound-model_disp)/sigma);
	double trunc_gaussian_part = truncation*gaussian_part;
	double lhood= (floor_proportion/(upper_bound-lower_bound) + (1-floor_proportion)*trunc_gaussian_part);
	if (measured_depth< 0){ // all images pixels with no range
	lhood = 1; // log(1) = 0 ---> has no effect
	}
	return log (lhood);
	}

[stale]

Marking this as stale due to 30 days of inactivity. It will be closed in 7 days if no further activity occurs.

[kunaltyagi]

@Morwenn We should close this issue, right? Based on the discussions before, using boost distributions provides better reproducibility.

[Morwenn]

Apparently the Boost components in RangeLikelihood have already been replaced by std:: ones a year ago, so the issue should have been closed regardless. If you want to rollback the distribution to the Boost one, I guess that it can be done as part of #3996 or a new issue.