Performs object detection with a multi-scale window

[rcts, weights] = hog.detectMultiScale(im)
[...] = hog.detectMultiScale(im, 'OptionName', optionValue, ...)

Input

• im 8-bit 1- or 3-channel image where objects are detected.

Output

• rcts Detected objects boundaries. Cell array of rectangles where objects are found, of the form {[x,y,w,h], ...}.
• weights Vector of associated weights. Contain confidence values for each detected object.

Options

• HitThreshold Threshold for the distance between features and SVM classifying plane. Usually it is 0 and should be specified in the detector coefficients (as the last free coefficient). But if the free coefficient is omitted (which is allowed), you can specify it manually here. default 0
• WinStride Window stride [w,h]. It must be a multiple of block stride. Not set by default in which case it uses BlockStride.
• Padding Padding [w,h]. default [0,0]
• Scale Coefficient of the detection window increase. default 1.05
• FinalThreshold Coefficient to regulate the similarity threshold. When detected, some objects can be covered by many rectangles. 0 means not to perform grouping. See cv.groupRectangles. default 2.0
• UseMeanshiftGrouping Flag to use meanshift grouping or the default grouping based on merging overlapped rectangles. When false, cv.HOGDescriptor.groupRectangles is performed and the value of FinalThreshold is used for GroupThreshold. When true, cv.groupRectangles_meanshift is performed instead, and FinalThreshold is used for the DetectThreshold option. default false

Detects objects of different sizes in the input image. The detected objects are returned as a list of rectangles.

SvmDetector should be set before calling this method.

The image is repeatedly scaled down by the specified Scale factor, as long as it remains larger than hog.WinSize or until a maximum of hog.NLevels levels is built. The resized images are then searched with a sliding window to detect objects similar to the cv.HOGDescriptor.detect method (this method is parallelized). Finally the found rectangles are grouped and clipped against the image size.