System and method for detecting generic items in image sequence

A merchandise checkout system for detecting a merchandise item from sequentially captured images, the system comprising:

• a feature extractor for extracting visual features from the sequentially captured images, the extracted visual features characterized by feature vectors;
• a feature comparison module operable to compare visual features extracted from a first image of the sequentially captured images to visual features extracted from a second image of the sequentially captured images to identify a group of extracted visual features that match between the first and second images;
• a classifier operable to measure, for the group of extracted visual features, a classification parameter so as to generate a parameter measurement that collectively characterizes the group of extracted visual features for comparison of the parameter measurement to an associated classification criterion, the associated classification criterion defining a first classification representing presence of merchandise items and a second classification representing presence of non-merchandise items, and the classifier further operable to classify the group of extracted visual features into either the first classification or the second classification based on whether the parameter measurement satisfies the associated classification criterion, wherein the associated classification criterion includes a threshold amount of entropy of an area occupied by the group of extracted visual features;
• and
• an alarm configured to generate an alert in response to the classifier classifying the group of extracted visual features as belonging to the first classification and thereby indicating the presence of the merchandise item.