Maybe an overflow bug in cascadedetect.cpp

[ddtuan99]

System information (version)

• OpenCV => 4.2
• Operating System / Platform => Windows 64 Bit
• Compiler => Visual Studio 2017

Detailed description

At this line of cascadedetect.cpp, the computeChannels method of HaarEvaluator computes integral image of input image and its squared pixel values with depth=CV_32S for both sum and sqsum. Also at this line, sbuf (which I belive is the buffer of the integral images) has type of CV_32S.

Even a 512x512 lena.jpg has sqsum exceed 32bits integer range, sum would soon be overflow for larger image. So I suspect that it would cause overflow bug thus make the whole code wrong except that the code was misunderstood (in this case, I apologize for my mistake).

Steps to reproduce

cascade.detectMultiScale()

Issue submission checklist

• I report the issue, it's not a question
• I checked the problem with documentation, FAQ, open issues,
  forum.opencv.org, Stack Overflow, etc and have not found solution
• I updated to latest OpenCV version and the issue is still there
• There is reproducer code and related data files: videos, images, onnx, etc