Photolitics

---

Photons Hit Electrons

---

Recent Pole-Zero Pairs

▨ First family of products

▨ Not deserted - temporarily stranded

▨ Analog-style digital design observations

▨ The semiconductor industry in Bulgaria Vol. 2

▨ Small-signal and large-signal MOSFET channel resistance

▨ Bricks and mortar of the internet

▨ Harvesting gamma radiation

▨ Dominant gate capacitance of a MOSFET

▨ Mr. Sievert, a fictional character

▨ "This could be a very strong paper"

---

• 
• 
• 
• 

Search

Chirp chirp...

---

Dilema

---

Deep N-Well

---

---

Simple motion detection with OpenCV

---

As a continuation of my previous post, here is a simple algorithm for motion detection in a live video stream using OpenCV. It basically follows a few simple steps:

1. Subtract frames and generate a binary image

cvAbsDiff(
                  frameTime1,
                  frameTime2,
                  frameForeground
                  );

        cvShowImage( "AbsDiff", frameForeground);  //AbsDiff window

        cvThreshold(
                frameForeground,
                frameForeground,
                20,  //Threshold
                255, //Saturate up to 255
                CV_THRESH_BINARY);
                 //CV_THRESH_BINARY_INV);
	         //CV_THRESH_TRUNC);
	         //CV_THRESH_TOZERO);
	         //CV_THRESH_TOZERO_INV);
	         //
        cvShowImage( "AbsDiffThreshold", frameForeground);  //AbsDiffThreshold window

The threshold as one may guess, can be used as a primitive noise supression parameter.

2. Run through the binary image and accumulate events.

	int row, col;
	uchar sig1, sig2;
	unsigned long int rowsum[frameForeground->height], totalsum;
	
	totalsum = 0;
	for( row = 0; row < frameForeground->height; row++ )
	{
	rowsum[row] = 0;
	  for ( col = 0; col < frameForeground->width; col++ )
	  {
	
	    sig1 = CV_IMAGE_ELEM( frameForeground, uchar, row, col * 2 );
	    sig2 = CV_IMAGE_ELEM( frameForeground, uchar, row, col * 2 + 1 );
	    rowsum[row] += (sig1 + sig2);

	    //printf("Y: %d X: %d Val: %d \n", row, col, sig1);
	    //printf("Y: %d X: %d Val: %d \n", row, col, sig2);
	
	  }

	totalsum += rowsum[row];
	
	}	

	printf("Totalsum: %20d \n", totalsum);

	if (totalsum >= 80000) // Motion detection threshold
	{ 
             cvRectangle(image,cvPoint(10,10), cvPoint(310, 230),cvScalar(0, 255, 0, 0),1,8,0); // Draw a green rectangle
	}

        cvShowImage( "Camera", image );  //Display the original image w/wo added green rectangle

After all event accumulation, a comparison with a motion detection coefficient is made. Once the total event count is larger than the detection coefficient a green rectangle is embedded on the original frame.

Here is my dead simple example, which should work straight out of the box.

Date:Tue Oct 26 15:33:31 BWT 2014

Except where otherwise noted, content on this site is copyright by Deyan Levski and available under a Creative Commons Attribution-NonCommercial 4.0 International License.