Object recognition system, object recognition method and object recognition program which are not susceptible to partial concealment of an object

An object recognition system for recognizing a category of an object based on an image of the object, comprising:

• a partial feature amount acquisition device which acquires a feature amount of a partial region of the image;
• a partial recognition device which recognizes the category in the partial region and obtains a partial recognition score;
• an occurrence probability calculation device which calculates an estimated value of an occurrence probability of the partial recognition score under a condition of being a recognition object category;
• a total score calculation device which calculates a total score based on nonoccurrence probabilities of the partial recognition scores of a plurality of the partial regions under the condition of being the recognition object category;
• and
• a result judgment device which judges the category of the object based on the total score,
• wherein the total score calculation device calculates the total score according to a numerical formula including an operation which divides a sum total of logarithms of the nonoccurrence probabilities of the partial recognition scores by a number of the partial regions.

An object recognition method for recognizing a category of an object based on an image of the object, comprising:

• executing, for each of the partial regions of the image, processing for acquiring a feature amount of a partial region of the image, processing for recognizing the category in the partial region and obtaining a partial recognition score, and processing for calculating an estimated value of an occurrence probability of the partial recognition score under a condition of being a recognition object category;
• calculating a total score based on nonoccurrence probabilities of the partial recognition scores of a plurality of the partial regions under the condition of being the recognition object category;
• and
• judging the category of the object based on the total score,
• wherein, when calculating the total score, the total score is calculated according to a numerical formula including a sum total of logarithms of the nonoccurrence probabilities of the partial recognition scores.

A non-transitory computer readable recording medium storing an object recognition program for recognizing a category of an object based on an image of the object, which enables a computer to execute:

• partial feature amount acquiring processing for acquiring a feature amount of a partial region of the image;
• partial recognition processing for recognizing the category in the partial region and obtaining a partial recognition score;
• occurrence probability calculation processing for calculating an estimated value of an occurrence probability of the partial recognition score under a condition of being a recognition object category;
• total score calculation processing for calculating a total score based on nonoccurrence probabilities of the partial recognition scores of a plurality of the partial regions under the condition of being the recognition object category;
• and
• result judgment processing for judging the category of the object based on the total score;
• and
• which enables the computer to calculate the total score according to a numerical formula including a sum total of logarithms of the nonoccurrence probabilities of the partial recognition scores.

An object recognition system for recognizing a category of an object based on an image of the object, comprising:

• partial feature amount acquisition means for acquiring a feature amount of a partial region of the image;
• partial recognition means for recognizing the category in the partial region and obtains a partial recognition score;
• occurrence probability calculation means for calculating an estimated value of an occurrence probability of the partial recognition score under a condition of being a recognition object category;
• total score calculation means for calculating a total score based on nonoccurrence probabilities of the partial recognition scores of a plurality of the partial regions under the condition of being the recognition object category;
• and
• result judgment means for judging the category of the object based on the total score,
• wherein the total score calculation device calculates the total score according to a numerical formula including a sum total of logarithms of the nonoccurrence probabilities of the partial recognition scores.

An object recognition system for recognizing a category of an object based on an image of the object, comprising:

• a partial feature amount acquisition means which acquires a feature amount of a partial region of the image;
• a partial recognition means which recognizes a category in the partial region and obtains a partial recognition score;
• an occurrence probability calculation means which calculates an estimated value of an occurrence probability of the partial recognition score under a condition of being a recognition object category;
• a total scare calculation means that calculates a total score based on nonoccurrence probabilities of the partial recognition scores of a plurality of the partial regions under the condition of being the recognition object category;
• and
• a result judgment means which judges the category of the object based on the total score,
• wherein the total score calculation means calculates the total score according to a numerical formula including an operation which divides a sum total of logarithms of the nonoccurrence probabilities of the partial recognition scores by a number of the partial regions.