Monitoring device

A monitoring device that detects a change in an object in a monitoring region from an input image acquired by a video acquisition unit, comprising:

• a reference image generation unit that generates a reference image as a reference for an image recognition process based on the input image;
• a foreground extraction unit that compares the input image with the reference image to detect a foreground region different from the reference image;
• a non-moving region extraction unit that compares the foreground region with the input image to extract a non-moving region from the foreground region;
• a feature amount extraction unit that extracts a feature amount of the non-moving region;
• a feature amount accumulation unit that accumulates the feature amount in a direction of time;
• a stationary object detection unit that performs clustering of pixels in which the feature amount accumulated by the feature amount accumulation unit satisfies a predetermined standard to detect a stationary object;
• a left object determination unit that determines whether the stationary object is a left object based on the feature amount of the stationary object and an evaluation of a surrounding environment;
• and
• a management unit that manages an attribute of the stationary object determined as the left object,
• wherein, as a means for indicating tendencies of standstill, congestion, and clogging of persons, for example, the feature amount accumulated image calculated by the feature accumulation unit is converted by an index of standstill degree, and the standstill degree is imaged and overlaid on the input image to visualize the tendencies of standstill.

A monitoring device that detects a change in an object in a monitoring region from an input image acquired by a video acquisition unit, comprising:

• a reference image generation unit that generates a reference image as a reference for an image recognition process based on the input image;
• a foreground extraction unit that compares the input image with the reference image to detect a foreground region different from the reference image;
• a non-moving region extraction unit that compares the foreground region with the input image to extract a non-moving region from the foreground region;
• a feature amount extraction unit that extracts a feature amount of the non-moving region;
• a feature amount accumulation unit that accumulates the feature amount in a direction of time;
• a stationary object detection unit that performs clustering of pixels in which the feature amount accumulated by the feature amount accumulation unit satisfies a predetermined standard to detect a stationary object;
• a left object determination unit that determines whether the stationary object is a left object based on the feature amount of the stationary object and an evaluation of a surrounding environment;
• and
• a management unit that manages an attribute of the stationary object determined as the left object,
• wherein an image in which a standstill degree heat map image acquired by performing spatial processing with a Gaussian filter on a feature amount accumulated image expressed in pseudo colors is overlaid on the input image by a blending is displayed.