People detection apparatus and method and people counting apparatus and method

A people counting method comprising:

• receiving a video of an area comprising an entrance;
• setting an inline at the entrance and setting an outline such that a specific region is formed on a side of the inline;
• detecting moving objects in the video using both a Gaussian mixture model (GMM) method and information differences between frames of the video, expanding moving object regions based on a priority determined from a difference in brightness between pixels of the moving object regions, and detecting human moving objects among the detected moving objects by using positions of the expanded moving object regions, sizes of the expanded moving object regions, and a histogram of oriented gradient (HOG), which is a shape feature descriptor;
• tracking movement of the detected moving objects;
• and
• determining, on an object basis, whether the moving objects each pass the inline and the outline based on the tracked movement, and counting the number of the people based on the determination result, with respect to ones of the moving objects detected as the human moving objects,
• wherein:
• the inline and the outline are virtual lines;
• and
• at least one of the receiving, setting, detecting, tracking, and determining is performed using a hardware processor.

A people detection method comprising:

• receiving a video frame of a video;
• calculating a respective accuracy value of detection for each pixel of the video frame using both a GMM method and a frame difference method;
• detecting, as moving object regions, pixels whose respective accuracy values are at least a preset accuracy value;
• expanding the moving object regions based on a priority determined from a difference in brightness between the pixels;
• and
• detecting human moving objects, among the moving objects, based on:
• positions of the moving object regions,
• sizes of the moving object regions, and
• a histogram of oriented gradient (HOG);
• wherein the HOG is a shape feature descriptor, and
• wherein the positions of the moving object regions are based on the expanding of the moving object regions;
• and
• wherein the sizes of the moving object regions are based on the expanding of the moving object regions.