Athena on November 23rd, 2011

When creating samples from the marked images the default options are listed below. You notice that the default for the maxzangle is less than the defaults for the x and y angle values. So why is that? What do these parameters mean?   The Haar Classifier was initially developed for facial detections, but has been […]

Continue reading about CreateSamples ::: [ Maxxangle, Maxyangle, Maxzangle ]

Athena on November 20th, 2011

In the previous post Advanced Object Marker ::: [Tracking defined ROI (Region Of Interest)] the second video shows the new region of interest (ROI) as the YELLOW rotated rectangle. For our training purposes, a rotated rectangle can not be input to the training application as it expects the defining members of a rectangle as x, […]

Continue reading about Bounding Rectangle for ROI of Rotated Rectangle

Now that we have eliminated scaling from our rotation images, we can now try to track/follow our region of interest (ROI) throughout all rotations. Below is the first initial attempt, we marked our first ROI so we know the left most point of the rectangle as well as its dimensions (width and height). So we […]

Continue reading about Advanced Object Marker ::: [Tracking defined ROI (Region Of Interest)]

So initially we thought to rotate the image 1 degree and save the resulting image, we then ended up with about 360 image files derived from a single image. If we marked just the initial image (non-rotated) then based off of the placement in that image we could derive the position of interest in all […]

Continue reading about Advanced Object Marker ::: [Rotating Images (Rotation and Scaling)]

Athena on October 20th, 2011

In order to simplify marking the positive images, instead of marking and identifying thousands of images we decided to try marking one image and rotating the image 1 degree for a complete 360 degree rotation. We are hoping that this will still bring about enough variance in our data sets, while still increasing the hit […]

Continue reading about Advanced Object Marker ::: [Cartesian Coordinate System]

Athena on October 3rd, 2011

So now that the limitation of the performance application has been identified what does that mean for the previous haar classifiers we have tested and analyzed this far? Well to put it bluntly they were wrong. The results did not truly reflect the classifiers accuracy. So I took the haarcascade15, since this classifier had the […]

Continue reading about Performance ::: haarcascade15 :::

Athena on October 2nd, 2011

When creating a classifier to use for detection of solar cavities, the steps taken are essentially creating the samples to train with, train, and then test the performance. In testing the performance, I began looking through the output, and noticed some oddity. For some of the images the performance output would classify a MISS, so […]

Continue reading about Performance Testing Accuracy

Athena on September 18th, 2011

OpenCV has functionality built in for template matching. The template matching implementation is basically matching a subimage against a main image by sliding it across the entire image using one of the matching methods, in my implementations I used the normalized square difference matching method. Below is a summarized excerpt for template matching within Quadrant […]

Continue reading about Template Matching for Solar Cavities

Athena on September 16th, 2011

Below are examples of the haar-like feature sets that are used in training. It displays what features are used based on the “mode” option used as well as the string representation that will be used in each of the AdaBoostCARTHaarClassifier.txt files created for each stage on the classifier. haar_x2 haar_y2 haar_x3 haar_y3 haar_x2_y2 haar_x4 haar_y4 […]

Continue reading about Haar-like Features

Athena on September 10th, 2011

In running through tests against all the haarcascades I create, I noticed that the number of Misses in Quadrant III seemed surprisingly high, even when the cavity was a very distinct one.   Here is an example of the original image that was used to train the cascades. Here is an output file from the […]

Continue reading about Missed Hits in Quadrant III