Information processing apparatus and image reading apparatus judging title of read document
In an image reading apparatus, a character area extractor extracts character areas from a document image in units of rows. A title reliability calculator calculates a title reliability level of each character area using a feature quantity data set and a machine learning model. A character recognizer converts character areas of which title reliability levels exceed a threshold into text data. A title judger collates text data with a title candidate. In a case in which one piece of text data coinciding with a title candidate is judged and detected, the title judger sets the text data as a title of the document image. In a case in which a plurality of pieces of coinciding text data are judged and detected, the title judger sets text data of which a title reliability level is the highest among the detected pieces of text data as a title of the document image.
1 . An information processing apparatus comprising a controller that includes a processor and, by the processor executing a control program, functions as:
a character area extractor configured to extract character areas from a document image acquired by reading a document in units of rows and acquire position information of each of the extracted character areas;
a feature quantity calculator configured to calculate a feature quantity of a character area using the position information of the character area and information of the document image;
a machine learning processor configured to generate a feature quantity data set configured using the feature quantities of the character areas and generate a machine learning model by performing machine learning of title judgment;
a title reliability calculator configured to calculate a title reliability level representing a possibility of inclusion of a title in a character area for each character area on the basis of the feature quantity data set and the machine learning model;
a character recognizer configured to perform character recognition of a character area having the title reliability level exceeding a threshold set in advance to convert the character area into text data;
a title candidate storage device configured to store character strings that are title candidates; and
a title judger configured to detect text data in which a character string stored by the title candidate storage device is included among text data converted by the character recognizer and judge the detected text data as a title of the document.
2 . The information processing apparatus according to claim 1 , wherein, in a case in which a plurality of pieces of text data including a character string stored by the title candidate storage device are judged, the title judger judges the text data of the character area having the highest title reliability level among the character areas represented by the text data as a title of the document.
3 . The information processing apparatus according to claim 1 , wherein, in a case in which the character recognizer judges that the character area having a title reliability level exceeding the threshold set in advance is absent or in a case in which the title judger judges that text data in which a character string stored by the title candidate storage device is included is absent in the text data converted by the character recognizer, the title judger judges that a title is not present in the document image.
4 . The information processing apparatus according to claim 1 , further comprising an input receiving device configured to receive an operation input,
wherein the input receiving device receives an input indicating whether or not a title is included in each character area included in the feature quantity data set, and
wherein the machine learning processor assigns a label indicating whether or not a title is included to each of the character areas of the feature quantity data set on the basis of details received by the input receiving device.
5 . The information processing apparatus according to claim 1 , wherein, when a label indicating that a title is included is assigned to one character area among the character areas, the machine learning processor assigns a label indicating that a title is included to this one character area and assigns a label indicating that no title is included to each of the character areas other than this one character area.
6 . The information processing apparatus according to claim 1 , wherein the feature quantity of the character area is determined using center coordinates of the character area in the document image and numerical values including a vertical length and a horizontal length of the character area.
7 . The information processing apparatus according to claim 1 , wherein the information of the document image is determined using numerical values including a vertical length and a horizontal length of the document image.
8 . An image reading apparatus comprising:
the information processing apparatus according to claim 1 ; and
an image reading device configured to acquire a document image by reading a document.