Assertion failed with accumulateWeighted in OpenCV - c++

Question

I am using openCV and trying to calculate a moving average of the background, then taking the current frame and subtracting the background to determine movement (of some sort).
However, when running the program I get:
OpenCV Error: Assertion failed (func != 0) in accumulateWeighted, file /home/sebbe/projekt/opencv/trunk/opencv/modules/imgproc/src/accum.cpp, line 431
terminate called after throwing an instance of 'cv::Exception'
what(): /home/sebbe/projekt/opencv/trunk/opencv/modules/imgproc/src/accum.cpp:431: error: (-215) func != 0 in function accumulateWeighted
I cant possibly see what arguments are wrong to accumulateWeighted.
Code inserted below:
#include <stdio.h>
#include <stdlib.h>
#include "cv.h"
#include "highgui.h"
#include "opencv2/imgproc/imgproc.hpp"
#include "opencv2/highgui/highgui.hpp"
#include "cxcore.h"
using namespace cv;
int main( int argc, char **argv )
{
Mat colourFrame;
Mat frame;
Mat greyFrame;
Mat movingAverage;
Mat difference;
Mat temp;
int key = 0;
VideoCapture cap(0);
/* always check */
if ( !cap.isOpened() ) {
fprintf( stderr, "Cannot open initialize webcam!\n" );
return 1;
}
namedWindow("Camera Window", 0);
// Initialize
cap >> movingAverage;
while( key != 'q' ) {
/* get a frame */
cap >> colourFrame;
/* Create a running average of the motion and convert the scale */
accumulateWeighted(colourFrame, movingAverage, 0.02, Mat() );
/* Take the difference from the current frame to the moving average */
absdiff(colourFrame, movingAverage, difference);
/* Convert the image to grayscale */
cvtColor(difference, greyFrame, CV_BGR2GRAY);
/* Convert the image to black and white */
threshold(greyFrame, greyFrame, 70, 255, CV_THRESH_BINARY);
/* display current frame */
imshow("Camera Window",greyFrame);
/* exit if user press 'q' */
key = cvWaitKey( 1 );
}
return 0;
}

Answer 1

Looking at the OpenCV sources, specifically at modules/imgproc/src/accum.cpp line 431, the lines that precede this assertion are:
void cv::accumulateWeighted( InputArray _src, CV_IN_OUT InputOutputArray _dst,
double alpha, InputArray _mask )
{
Mat src = _src.getMat(), dst = _dst.getMat(), mask = _mask.getMat();
int sdepth = src.depth(), ddepth = dst.depth(), cn = src.channels();
CV_Assert( dst.size == src.size && dst.channels() == cn );
CV_Assert( mask.empty() || (mask.size == src.size && mask.type() == CV_8U) );
intfidx = getAccTabIdx(sdepth, ddepth);
AccWFunc func = fidx >= 0 ? accWTab[fidx] : 0;
CV_Assert( func != 0 ); // line 431
What's happening in your case is that getAccTabIdx() is returning -1, which in turn makes func be ZERO.
For accumulateWeighted() to work properly, the depth of colourFrame and movingAverage must be one of the following options:
colourFrame.depth() == CV_8U && movingAverage.depth() == CV_32F
colourFrame.depth() == CV_8U && movingAverage.depth() == CV_64F
colourFrame.depth() == CV_16U && movingAverage.depth() == CV_32F
colourFrame.depth() == CV_16U && movingAverage.depth() == CV_64F
colourFrame.depth() == CV_32F && movingAverage.depth() == CV_32F
colourFrame.depth() == CV_32F && movingAverage.depth() == CV_64F
colourFrame.depth() == CV_64F && movingAverage.depth() == CV_64F
Anything different than that will make getAccTabIdx() return -1 and trigger the exception at line 431.

Answer 2

From the documentation on OpenCV API you can see that the output image from accumulateWeighted is
dst – Accumulator image with the same number of channels as input image, 32-bit or 64-bit floating-point.
So your initialization is wrong. You should retrieve the colourFrame size first and then do this:
cv::Mat movingAverage = cv::Mat::zeros(colourFrame.size(), CV_32FC3);

Answer 3

On Python a working solution is to initiate movingAverage using FIRSTcolourFrame.copy().astype("float").
I found the solution on this website