Method and apparatus for extracting and processing price text from an image set

Abstract

A method of extracting price text from an image set includes: obtaining input data comprising (i) a plurality of images depicting shelves supporting products, and (ii) for each of the images, a set of text regions and corresponding price text strings; registering the images to a common frame of reference; identifying a subset of the text regions having overlapping locations in the common frame of reference; selecting one of the text regions from the subset; and presenting the price text string corresponding to the one of the text regions for further processing.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS

This application is related to U.S. patent application Ser. No. 15/583,786, entitled “METHOD AND APPARATUS FOR LABEL DETECTION” by Lam, as well as U.S. Provisional Patent Application No. 62/492,670, entitled “PRODUCT STATUS DETECTION SYSTEM” by Perrella et al., all having the same filing date as the present application. The contents of the above-reference applications are incorporated herein by reference in their entirety.

BACKGROUND

Environments in which inventories of objects are managed, such as products for purchase in a retail environment, may be complex and fluid. For example, a given environment may contain a wide variety of objects with different attributes (size, shape, price and the like). Further, the placement and quantity of the objects in the environment may change frequently. Still further, imaging conditions such as lighting may be variable both over time and at different locations in the environment. These factors may reduce the accuracy with which information concerning the objects may be collected within the environment.

BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWINGS

The accompanying figures, where like reference numerals refer to identical or functionally similar elements throughout the separate views, together with the detailed description below, are incorporated in and form part of the specification, and serve to further illustrate embodiments of concepts that include the claimed invention, and explain various principles and advantages of those embodiments.

FIG. 1 is a schematic of a mobile automation system.

FIG. 2 is a block diagram of certain internal hardware components of the server in the system of FIG. 1.

FIG. 3 is a flowchart of a method of price text detection and interpretation.

FIG. 4 depicts a set of images obtained during the performance of the method of FIG. 3.

FIG. 5 depicts the text regions of the images of FIG. 4 in detail.

FIG. 6 depicts the images of FIG. 4 following registration.

FIG. 7 depicts a method of identifying text region subsets during the performance of the method of FIG. 3.

FIG. 8 illustrates the performance of the method of FIG. 7.

Skilled artisans will appreciate that elements in the figures are illustrated for simplicity and clarity and have not necessarily been drawn to scale. For example, the dimensions of some of the elements in the figures may be exaggerated relative to other elements to help to improve understanding of embodiments of the present invention.

The apparatus and method components have been represented where appropriate by conventional symbols in the drawings, showing only those specific details that are pertinent to understanding the embodiments of the present invention so as not to obscure the disclosure with details that will be readily apparent to those of ordinary skill in the art having the benefit of the description herein.

DETAILED DESCRIPTION

Environments such as warehouses, retail locations (e.g. grocery stores) and the like typically contain a wide variety of products supported on shelves, for selection and purchase by customers. The products are generally labelled—for example, via a label placed on a shelf edge, a peg extending from a back of the shelf, or directly on the product itself—with information such as the price of the product, an identifier such as a SKU number, and the like. Such environments typically also store reference data relating to the products, for example in a central database, which is consulted by point-of-sale terminals during customer checkout to retrieve price information for the products being purchased. In some cases, the price physically labelled on or near the product on the shelves may not match the price stored in the above-mentioned database, leading to a conflict between the label price and the reference price at the point-of-sale terminal.

Mismatches between label and reference prices may require corrective action at one or both of the label and the central database. However, detecting such mismatches in order to allow corrective action to be taken is conventionally performed by human employees, via visual assessment of the shelves and manually-initiated barcode scanning. This form of detection is labor-intensive and therefore costly, as well as error-prone.

Attempts to automate the detection of such mismatches, for example via machine vision techniques, are complicated by difficulties in correctly extracting labelled prices from image data. Such difficulties include non-label text present on products, as well as imaging artifacts such as variations in contrast, reflections, and the like. Autonomous detection of mismatches between label information and reference information is further complicated by the need to retrieve the correct reference data with which to compare the price text extracted from shelf images. To retrieve reference data, the identity of the product to which the label text relates must also be autonomously detected—a process that is subject to similar difficulties as those outlined above in connection with text detection and interpretation.

Examples disclosed herein are directed to a method of extracting price text from an image set, including: obtaining input data comprising (i) a plurality of images depicting shelves supporting products, and (ii) for each of the images, a set of text regions and corresponding price text strings; registering the images to a common frame of reference; identifying a subset of the text regions having overlapping locations in the common frame of reference; selecting one of the text regions from the subset; and presenting the price text string corresponding to the one of the text regions for further processing.

FIG. 1 depicts a mobile automation system 100 in accordance with the teachings of this disclosure. The system 100 includes a server 101 in communication with at least one mobile automation apparatus 103 (also referred to herein simply as the apparatus 103) and at least one mobile device 105 via communication links 107, illustrated in the present example as including wireless links. The system 100 is deployed, in the illustrated example, in a retail environment including a plurality of shelf modules 110 each supporting a plurality of products 112. The shelf modules 110 are typically arranged in a plurality of aisles, each of which includes a plurality of modules aligned end-to-end. More specifically, the apparatus 103 is deployed within the retail environment, and communicates with the server 101 (via the link 107) to navigate, either fully or partially autonomously, the length of at least a portion of the shelves 110. The apparatus 103 is equipped with a plurality of navigation and data capture sensors 104, such as image sensors (e.g. one or more digital cameras) and depth sensors (e.g. one or more Light Detection and Ranging (LIDAR) sensors), and is further configured to employ the sensors to capture shelf data. In the present example, the apparatus 103 is configured to capture a series of digital images of the shelves 110, as well as a series of depth measurements, each describing the distance and direction between the apparatus 103 and one or more points on a shelf 110, such as the shelf itself or the product disposed on the shelf.

The server 101 includes a special purpose imaging controller, such as a processor 120, specifically designed to control the mobile automation apparatus 103 to capture data, obtain the captured data via the communications interface 124 and store the captured data in a repository 132 in the memory 122. The server 101 is further configured to perform various post-processing operations on the captured data and to detect the status of the products 112 on the shelves 110. When certain status indicators are detected by the imaging processor 120, the server 101 is also configured to transmit status notifications (e.g. notifications indicating that products are out-of-stock, low stock or misplaced) to the mobile device 105. The processor 120 is interconnected with a non-transitory computer readable storage medium, such as a memory 122, having stored thereon computer readable instructions for executing price text detection and interpretation, as discussed in further detail below. The memory 122 includes a combination of volatile (e.g. Random Access Memory or RAM) and non-volatile memory (e.g. read only memory or ROM, Electrically Erasable Programmable Read Only Memory or EEPROM, flash memory). The processor 120 and the memory 122 each comprise one or more integrated circuits. In an embodiment, the processor 120, further includes one or more central processing units (CPUs) and/or graphics processing units (GPUs). In an embodiment, a specially designed integrated circuit, such as a Field Programmable Gate Array (FPGA), is designed to perform the price text detection and interpretation discussed herein, either alternatively or in addition to the imaging controller/processor 120 and memory 122. As those of skill in the art will realize, the mobile automation apparatus 103 also includes one or more controllers or processors and/or FPGAs, in communication with the controller 120, specifically configured to control navigational and/or data capture aspects of the apparatus 103.

The server 101 also includes a communications interface 124 interconnected with the processor 120. The communications interface 124 includes suitable hardware (e.g. transmitters, receivers, network interface controllers and the like) allowing the server 101 to communicate with other computing devices—particularly the apparatus 103 and the mobile device 105—via the links 107. The links 107 may be direct links, or links that traverse one or more networks, including both local and wide-area networks. The specific components of the communications interface 124 are selected based on the type of network or other links that the server 101 is required to communicate over. In the present example, a wireless local-area network is implemented within the retail environment via the deployment of one or more wireless access points. The links 107 therefore include both wireless links between the apparatus 103 and the mobile device 105 and the above-mentioned access points, and a wired link (e.g. an Ethernet-based link) between the server 101 and the access point.

The memory 122 stores a plurality of applications, each including a plurality of computer readable instructions executable by the processor 120. The execution of the above-mentioned instructions by the processor 120 configures the server 101 to perform various actions discussed herein. The applications stored in the memory 122 include a control application 128, which may also be implemented as a suite of logically distinct applications. In general, via execution of the control application 128 or subcomponents thereof, the processor 120 is configured to implement various functionality. The processor 120, as configured via the execution of the control application 128, is also referred to herein as the controller 120. As will now be apparent, some or all of the functionality implemented by the controller 120 described below may also be performed by preconfigured hardware elements (e.g. one or more ASICs) rather than by execution of the control application 128 by the processor 120.

In the present example, in particular, the server 101 is configured via the execution of the control application 128 by the processor 120, to process image and depth data captured by the apparatus 103 to extract price text strings from sets of shelf images, and to identify and retrieve corresponding reference data for comparison to the price text strings.

Turning now to FIG. 2, before describing the operation of the application 128 to identify and interpret price text from captured image data, certain components of the application 128 will be described in greater detail. As will be apparent to those skilled in the art, in other examples the components of the application 128 may be separated into distinct applications, or combined into other sets of components. Some or all of the components illustrated in FIG. 2 may also be implemented as dedicated hardware components, such as one or more Application-Specific Integrated Circuits (ASICs) or FPGAs. For example, in one embodiment, to improve reliability and processing speed, at least some of the components of FIG. 2 are programmed directly into the imaging controller 120, which may be an FPGA or an ASIC having circuit and memory configuration specifically designed to optimize image processing and detection of high volume price label image data being received from the mobile automation apparatus 103. In such an embodiment, some or all of the control application 128, discussed below, is an FPGA or an ASIC chip.

The control application 128, in brief, includes components configured to obtain a plurality of shelf images and, with each image, a set of text region locations and corresponding price text strings. The components of the control application 128 are further configured to identify certain subsets of the above-mentioned text regions and select one text region from each subset. Further, the components of the control application 128 are configured to obtain reference data corresponding to the selected text region, and to compare the price string of the selected region with the reference data to detect mismatches.

More specifically, in the present example, the control application 128 includes an image set processor 200 configured to obtain and register the plurality of shelf images. The control application 128 further includes a selector 204 configured to identify subsets of text regions and select a text region from each subset, as well as an extractor 208 configured to extract portions of the images corresponding to the selected text regions, for use in obtaining reference data. In addition, the control application 128 includes a comparator 212 configured to obtain reference data employing the output of the extractor 208, and to determine whether a mismatch exists between the reference data and the price text associated with the images.

The functionality of the control application 128 will now be described in greater detail, with reference to the components illustrated in FIG. 2. Turning to FIG. 3, a method 300 of price text extraction and processing is shown. The method 300 will be described in conjunction with its performance on the system 100 as described above.

The performance of the method 300 begins at block 305, at which the controller 120, and in particular the image set processor 200, is configured to obtain and register a set of shelf images, such as digital images of the shelf 110, for example captured by the apparatus 103 and stored in the repository 132. An example set of images 400-1, 400-2 and 400-3 is illustrated in FIG. 4, depicting successive portions of a shelf 110 having a shelf edge 404 and a shelf back 408, as well as a support surface 412 extending the between the shelf edge 404 and the shelf back 408 and supporting products 112. In the present example, the set of images 400 is acquired by the apparatus 103 as the apparatus 103 moves among the shelves 110 shown in FIG. 1. Therefore, each portion of a shelf module is typically depicted in more than one image. For example, a product 112-1 is depicted in the images 400-1 and 400-2, while a product 112-2 is depicted in all three images 400 shown in FIG. 4, and a product 112-3 is depicted in the images 400-2 and 400-3.

At block 305, the image set processor 200 is also configured to obtain, for each of the images 400, a set of text regions and corresponding price text strings. The text regions correspond to regions within the images 400 that have been determined to contain price text, for example by another component of the control application 128. In brief, each text region indicates an area within an image 400 that likely contains a price text string, as well as a machine-readable interpretation of the price text string (generated, for example, by an optical character recognition or OCR operation). In other words, each text region indicates an area within an image 400 that likely depicts a price label corresponding to one of the products 112.

The set of text regions and price text strings may be obtained at block 305 in a variety of formats. For example, the text regions and price text strings may be indicated in the image data itself, as an overlay or in one or more metadata fields. In the present example, the text regions and price text strings are obtained from the repository 132 as a list of bounding box definitions and corresponding strings, as shown below in Table 1. The bounding boxes 416 are illustrated in FIG. 4 and identified by the text region identifiers shown in Table 1 for illustrative purposes. In other examples, identifiers need not be assigned to the text regions, and the first column of Table 1 can therefore be omitted.

TABLE 1
Text Regions and Price Text Strings
			Price
	Image/Text		Text	Confidence
	Region ID	Bounding Box	String	Level
	400-1/416-0	[X1, Y1] [X2, Y2]	$10.89	72%
	400-1/416-1	[X1, Y1] [X2, Y2]	$2.45	91%
	400-1/416-2	[X1, Y1] [X2, Y2]	$12.98	84%
	400-2/416-0	[X1, Y1] [X2, Y2]	$10.88	85%
	400-2/416-1	[X1, Y1] [X2, Y2]	$2.45	83%
	400-2/416-2	[X1, Y1] [X2, Y2]	$12.99	92%
	400-2/416-3	[X1, Y1] [X2, Y2]	$45.15	89%
	400-3/416-0	[X1, Y1] [X2, Y2]	$18.99	87%
	400-3/416-2	[X1, Y1] [X2, Y2]	$12.09	90%
	400-3/416-3	[X1, Y1] [X2, Y2]	$45.16	78%

The bounding box coordinates shown in Table 1 are typically indicated as pixel positions within each image; in the example above, two pixel positions are included in each record, corresponding to the top-left and bottom-right corners of each bounding box. As will now be apparent, a variety of other formats may be employed for storing the bounding box locations and dimensions.

As seen in Table 1, the data obtained at block 305 also includes the above-mentioned price text strings and, in association with each price text string, a confidence level value. The confidence levels shown above are formatted as percentages, but any other suitable format may also be used to convey a level of certainty that the price text string correctly reflects the price text printed on the labels depicted in the images 400. The confidence levels are typically generated during the OCR operation mentioned earlier, employed to generate the price text strings from image data.

Turning to FIG. 5, each of the text regions 416 is illustrated in greater detail, in isolation from the remainder of the images 400. As will be apparent, the price text strings in Table 1 do not always match the price text displayed on the labels depicted by the images 400 within the regions 416. Mismatches may result, for example, from imaging artifacts that can lead to the incorrect OCR interpretation of certain characters. As will also be apparent, each physical label depicted in the images 400 is represented a plurality of times in the data obtained at block 305 (as a result of appearing in multiple images). The confidence levels mentioned above, together with the multiple representations of each label in the images 400, are employed in the method 300 to select among the text regions 416.

Referring to FIG. 6, the images 400-1, 400-2 and 400-3 are shown following their registration to a common frame of reference by the image set processor 200. The common frame of reference is a single set of coordinates in which each of the pixels of the images 400 is assigned a position. The common frame of reference is specific to the images 400 in some examples; in other examples, the common frame of reference coordinates correspond to locations in space within the retail environment. Various image registration techniques may be employed at block 305, including feature-based techniques (e.g. edge and blob identification and matching), intensity-based techniques, or a combination thereof. In other examples, the image registration is based on, or supplemented with, navigational data obtained from the apparatus 103 defining the position and orientation of the apparatus 103 within the retail environment when each of the images 400 was captured. As seen in FIG. 6, some features of the images 400 do not align perfectly following registration. This may result from, for example, the fact that each image is captured at a different angle relative to the shelf 110 as the apparatus 103 travels the aisles between shelves 110. However, even for features that do not perfectly register, it is plain from FIG. 6 that some features of the images, including the portions of the images 400 corresponding to the text regions 416, overlap to varying degrees.

Returning to FIG. 3, at block 310 the selector 204 is configured to identify one or more subsets of the text regions 416 having overlapping locations in the common frame of reference to which the images 400 were registered at block 305. Referring again to FIG. 6, four subsets of text regions 416 are illustrated, labelled as subset 600-0 (including overlapping depictions of the text region 416-0), subset 600-1 (including overlapping depictions of the text region 416-1), subset 600-2 (including overlapping depictions of the text region 416-2), and subset 600-3 (including overlapping depictions of the text region 416-3). Thus, at block 310 the selector 204 is configured, in the present example, to identify each of the subsets 600 for further processing.

In other examples, the selector 204 is configured to perform additional actions at block 310. Referring to FIG. 7, a method of identifying subsets of text regions is illustrated. At block 705, the selector 204 is configured to obtain a shelf edge location. The shelf edge location may be retrieved, for example, from the repository 132, by querying the repository for a shelf edge location at a given position within the retail environment (e.g. defined by navigational data associated with the images 400). The selector 204 is also configured to register the shelf edge location with the registered set of images shown in FIG. 6. For example, the shelf edge location is stored in the repository as a height from a ground or other reference surface, as well as a location within the retail environment. The location and height of the shelf edge may be mapped to the registered image set 600 by any suitable registration technique. As a result of such registration, the shelf edge location may be overlaid on the registered set of images obtained at block 305 of the method 300.

At block 710, the selector 204 is configured to perform a shelf edge validation. The image data obtained at block 305 typically also includes depth measurements obtained by the apparatus 103 and stored in the repository 132 in association with the images 400. The depth measurements can therefore also be registered with the images 400, such that at least a subset of the pixels in each image 400 is mapped to a depth measurement, representing a location in a point cloud.

The selector 204 is configured to validate the retrieved shelf edge location by cropping the point cloud mentioned above to select a subset of the points therein falling within the retrieved location of the shelf edge. In other words, the depth measurements that do not fall within the retrieved shelf edge location are discarded or ignored for the purposes of block 710. Having cropped the point cloud, the selector 204 is then configured to group the remaining points into clusters based on their relative positions. In the present example, the selector 204 is configured to compare each pair of points in the cropped point cloud, and when the distance between the pair is below a predetermined threshold, the pair of points are added to a cluster. This process is repeated until all points in the cropped point cloud have been assigned to clusters. Other cluster-building operations may also be performed, as will be apparent to those skilled in the art.

The selector 204 is then configured to select a primary one of the clusters of points for further processing. More specifically, the selector 204 is configured, in the present example, to select the largest cluster, as defined by the number of points in the cluster. The selector 204 can also apply a minimum size threshold, and thus select the largest cluster that meets the minimum size threshold. If no clusters meet the minimum size threshold, performance of the method 315 proceeds to block 720, as will be discussed earlier.

Having selected a primary cluster of points, the selector 204 is configured to fit a plane to the cluster of points by a suitable plane-fitting operation (e.g. the random sample consensus, or RANSAC, plane fitting operation; or a least squares fitting operation, in which the square of the distance between each point and the plane is minimized). The selector 204 is further configured to determine the normal of the fitted plane. That is, the selector 204 is configured to determine the orientation of a line or vector that is perpendicular to the fitted plane. The selector 204 is further configured to determine whether the orientation of the normal matches a predetermined target orientation corresponding to the normal of the known shelf edge. For example, given that shelf edges are typically vertical surfaces, the target orientation may be a horizontal orientation (i.e. substantially parallel to a ground surface in the retail environment). The repository 132 may specify any suitable target orientation, however, based on the physical structure of the shelf edge 404. When the orientation of the normal substantially matches the target orientation, the retrieved shelf edge location is presumed to be valid. That is, the selector 204 is configured to assume that the plane fitted to the selected cluster of points corresponds to a portion of the shelf edge 404, because the orientation of that plane matches the expected orientation of the shelf edge 404. In such a situation, the determination at block 715 is therefore affirmative and the performance of the method 315 proceeds to block 725.

When the determination at block 715 is negative, performance of the method 315 instead proceeds to block 720. Block 720 is also performed when no shelf location is available from the repository 132, or when, as noted above, no cluster of depth measurements exceeds a minimum size threshold. At block 720, the selector 204 is configured to request an updated shelf edge location. The request can be generated and sent, for example, to another component of the control application 128 which performs a shelf edge detection operation employing the images 400 and depth measurements.

At block 725, having either validated the shelf edge location obtained at block 705 or received a shelf edge location at block 720, the selector 204 is configured to discard any text regions 416 that are not coincident with the shelf edge location. Referring to FIG. 8, a shelf edge location bounding box 800 is shown overlaid on the registered set of images 400. As is evident from FIG. 8, the text regions 416-0 from each of the images 400-1, 400-2 and 400-3 are not coincident with the shelf edge location 800. Therefore, in the present example the text regions 416-0 are ignored for further analysis. As will now be apparent, in some cases a text region 416 may partially overlap with the shelf edge location 800. The selector 204 is configured, in some examples, to determine a degree of overlap between the shelf edge location 800 and a text region 416 (e.g. how much of the area of the text region 416 overlaps with the shelf edge location 800), and determine whether the degree of overlap exceeds a threshold. If the degree of overlap does not exceed the threshold, the text region 416 is discarded.

At block 730, the selector 204 is configured to identify the subsets of overlapping text regions 416 from the text regions remaining after the performance of block 725. Thus, in the present example, referring to FIGS. 6 and 8, the subsets 600-1, 600-2 and 600-3, which are coincident with the shelf edge location 800, are selected at block 730.

Returning to FIG. 3, at block 315 the selector 204 is configured to select a text region 416 from each of the subsets identified at block 310. In the present example, the selector 204 is configured to make the selection at block 315 by comparing the confidence levels corresponding to each of the subset of text regions 416 under consideration, and selecting the text region 416 having the highest confidence level. Therefore, in the present example performance of the method 300, as is evident from Table 1, the text region 416-1 from the image 400-1 is selected from the text regions of the subset 600-1. Further, the text region 416-2 from the image 400-2 is selected from the subset 600-2, and the text region 416-3 from the image 400-2 is selected from the subset 600-3. As will be discussed below in further detail, the selection of text regions at block 315 also selects the corresponding price text strings for later use.

At block 320, the extractor 208 is configured to extract an area of at least one of the images 400 corresponding to each identified subset 600. In the present example, the area is extracted from the registered image set shown in FIG. 6. In other examples, however, the area is extracted from one of the images 400. The area extracted for a given subset 600 corresponds to the area of the image encompassed by the text regions 416 in the subset. In other words, the extractor 208 is configured to extract areas of the images 400 corresponding to the text regions 416 illustrated in FIG. 6. The appearance of the extracted areas is therefore as shown in FIG. 5 (with the exception of the text region 416-0, which was excluded as discussed earlier).

As seen in FIG. 5, each of the text regions 416 (and therefore each of the extracted image areas) depicts a barcode. The detection of the text regions 416 typically defines the text regions 416 so as to include nearby barcodes (e.g. as the barcodes themselves may be identified as potential text elements). In some examples, the extractor is configured to extract areas that are larger than the text regions 416 by a predefined padding margin (e.g. a number of pixels on each side of the text regions 416).

Following the extraction of the image areas, the extractor 208 is configured to obtain reference data by providing each extracted image area to a barcode decoder component of the control application 128 (not shown), or of any other element of the system 100. Any suitable decoder may be implemented to identify and decode the barcodes in the extracted areas of the images 400. The extractor 208 is therefore also configured to receive, from the above-mentioned decoder, decoded data corresponding to each extracted image area, such as product identifiers (e.g. UPC numbers). In other examples, rather than extracting areas of the image for transmission to the decoder, the extractor is configured to transmit the entire registered set of images 400 to the decoder, and receive a plurality of decoded data strings along with indications of their locations within the images. In such examples, the extractor is configured to associate each selected text region 416 with one of the barcodes, based on distances between the barcodes and the text region 416. For example, the extractor 208 can be configured to determine the distance between each pair of a text region and a barcode, and associate with each text region 416 the closest barcode to that text region 416.

In further examples, such as those in which the labels do not include barcodes, the reference data is obtained not by decoding barcodes but by providing the images 400 to a product recognition engine, which may be a further component of the control application 128, or a separate component from the control application 128. The product recognition engine is configured to compare the images 400 to a database of product models and, for example based on image feature matching, identify products within the image data and return product identifiers and their locations in the image, which are employed to retrieve a reference price. At block 320, the comparator 212 is therefore configured, in such examples, to receive from the product recognition engine a set of product locations within the images 400 and corresponding product identifiers. The comparator 212 is also configured to associate the received product identifiers with the price text strings, for example based on the distance between the product location (within the image 400) received from the product recognition engine and the location of the price text string within the image 400. In further examples, geometric constraints may be applied to the above association, in addition to an evaluation of distance between price text and product. For example, a price text string may be associated with a product identifier if the price text string was obtained from an area of the image 400 that is both within a threshold distance of the product location and below or above, rather than to one side of, the product location.

Having obtained decoded data from the extracted areas of the images 400, the extractor 208 is configured to then transmit a request, for example to the repository 132, containing the decoded data (or the product identifier received from the product recognition engine) and requesting a reference price for each item of decoded data (that is, for each of the selected text regions 416). As noted earlier, the repository 132 or any other suitable data source stores reference data for the products 112, including product identifiers and corresponding reference price strings. The extractor 208 is configured, upon receiving the reference prices, to pass the product identifiers obtained from the barcodes, the price text strings obtained at block 305 for the selected text regions 416, and the reference price strings obtained at block 320, to the comparator 312. An example of data passed to the comparator 312 is shown in Table 2.

TABLE 2
Comparator Inputs
Image/Text
Region ID	Price Text String	Product ID	Reference Price
400-1/416-1	$2.45	12345	$2.45
400-2/416-2	$12.99	12346	$12.20
400-2/416-3	$45.15	12347	$45.15

The comparator 212, responsive to receiving the price text strings and reference prices, is configured at block 325 to determine whether the price text string matches the corresponding reference price for each selected text region 416. As will be apparent from Table 2, the price text string and reference price for the text regions 416-1 and 416-3 match, while the price text string and reference price for the text region 416-2 do not match. Therefore, the performance of the method 300 proceeds to block 330, at which the comparator 212 is configured to generate and present a mismatch alert, for example via a user interface of the mobile device 105 or another remote computing device. The mismatch alert generated at block 330 includes the product identifier (e.g. as shown in Table 2), as well as the location within the retail environment that corresponds to the text region 416. For example, the location may indicate the aisle and shelf module on which the label corresponding to the mismatched text region 416 is located. The mismatch alert can be presented in a variety of ways. In some examples, the alert is transmitted from the server 101 to the mobile device 105 for presentation on a display of the mobile device 105. In other examples, the alert is stored in the repository 132 for further processing.

When the determination at block 325 is affirmative, the performance of the method 300 proceeds to block 335, at which the match result is presented, for example by storing an indication in the repository 132 that a successful price compliance check has been performed for the relevant product identifier. In other examples, block 330 may be omitted.

In the foregoing specification, specific embodiments have been described. However, one of ordinary skill in the art appreciates that various modifications and changes can be made without departing from the scope of the invention as set forth in the claims below. Accordingly, the specification and figures are to be regarded in an illustrative rather than a restrictive sense, and all such modifications are intended to be included within the scope of present teachings.

The benefits, advantages, solutions to problems, and any element(s) that may cause any benefit, advantage, or solution to occur or become more pronounced are not to be construed as a critical, required, or essential features or elements of any or all the claims. The invention is defined solely by the appended claims including any amendments made during the pendency of this application and all equivalents of those claims as issued.

Moreover in this document, relational terms such as first and second, top and bottom, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. The terms “comprises,” “comprising,” “has”, “having,” “includes”, “including,” “contains”, “containing” or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises, has, includes, contains a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by “comprises . . . a”, “has . . . a”, “includes . . . a”, “contains . . . a” does not, without more constraints, preclude the existence of additional identical elements in the process, method, article, or apparatus that comprises, has, includes, contains the element. The terms “a” and “an” are defined as one or more unless explicitly stated otherwise herein. The terms “substantially”, “essentially”, “approximately”, “about” or any other version thereof, are defined as being close to as understood by one of ordinary skill in the art, and in one non-limiting embodiment the term is defined to be within 10%, in another embodiment within 5%, in another embodiment within 1% and in another embodiment within 0.5%. The term “coupled” as used herein is defined as connected, although not necessarily directly and not necessarily mechanically. A device or structure that is “configured” in a certain way is configured in at least that way, but may also be configured in ways that are not listed.

It will be appreciated that some embodiments may be comprised of one or more generic or specialized processors (or “processing devices”) such as microprocessors, digital signal processors, customized processors and field programmable gate arrays (FPGAs) and unique stored program instructions (including both software and firmware) that control the one or more processors to implement, in conjunction with certain non-processor circuits, some, most, or all of the functions of the method and/or apparatus described herein. Alternatively, some or all functions could be implemented by a state machine that has no stored program instructions, or in one or more application specific integrated circuits (ASICs), in which each function or some combinations of certain of the functions are implemented as custom logic. Of course, a combination of the two approaches could be used.

Moreover, an embodiment can be implemented as a computer-readable storage medium having computer readable code stored thereon for programming a computer (e.g., comprising a processor) to perform a method as described and claimed herein. Examples of such computer-readable storage mediums include, but are not limited to, a hard disk, a CD-ROM, an optical storage device, a magnetic storage device, a ROM (Read Only Memory), a PROM (Programmable Read Only Memory), an EPROM (Erasable Programmable Read Only Memory), an EEPROM (Electrically Erasable Programmable Read Only Memory) and a Flash memory. Further, it is expected that one of ordinary skill, notwithstanding possibly significant effort and many design choices motivated by, for example, available time, current technology, and economic considerations, when guided by the concepts and principles disclosed herein will be readily capable of generating such software instructions and programs and ICs with minimal experimentation.

The Abstract of the Disclosure is provided to allow the reader to quickly ascertain the nature of the technical disclosure. It is submitted with the understanding that it will not be used to interpret or limit the scope or meaning of the claims. In addition, in the foregoing Detailed Description, it can be seen that various features are grouped together in various embodiments for the purpose of streamlining the disclosure. This method of disclosure is not to be interpreted as reflecting an intention that the claimed embodiments require more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive subject matter lies in less than all features of a single disclosed embodiment. Thus the following claims are hereby incorporated into the Detailed Description, with each claim standing on its own as a separately claimed subject matter.

Claims

1. A method of extracting price text from an image set, comprising: obtaining, from an image sensor by an image set processor, input data comprising (i) a plurality of images depicting shelves supporting products, and (ii) for each of the images, a set of text regions and corresponding price text strings, wherein each text region is indicated by a set of pixel coordinates within the corresponding image obtained from the image sensor;registering, by the image set processor, the images to a common frame of reference;identifying, by a selector interconnected with the image set processor, a subset of the text regions having overlapping locations in the common frame of reference by retrieving from a repository a shelf edge location in the common frame of reference based on navigational data associated with each of the plurality of images and discarding any text regions that are not coincident with the shelf edge location;selecting, by the selector, one of the text regions from the subset; andpresenting, by the selector, the price text string corresponding to the one of the text regions for comparison to a reference price.

2. The method of claim 1, wherein identifying the subset of text regions comprises identifying text regions having at least a predefined degree of overlap in the common frame of reference.

3. The method of claim 1, wherein the input data further comprises a confidence value corresponding to each of the price text strings.

4. The method of claim 3, wherein selecting one of the text regions comprises selecting the one of the subset of text regions having the highest confidence value.

5. The method of claim 1, further comprising: obtaining a plurality, of depth measurements corresponding to the plurality of images; and responsive to obtaining the shelf edge location, performing a validation of the shelf edge location based on the depth measurements.

6. The method of claim 5, where performing the shelf edge validation comprises: selecting a subset of the depth measurements falling within the shelf edge location obtained in the common frame of reference;clustering the subset of the depth measurements into a plurality of clusters;selecting a primary one of the clusters;fitting a plane to the primary cluster; anddetermining whether the plane has a preconfigured orientation.

7. The method of claim 1, further comprising: obtaining a reference price;comparing the price text string corresponding to the one of the subset of text regions to the reference price; andpresenting a result of the comparison.

8. The method of claim 7, wherein obtaining the reference price comprises: extracting an area of at least one of the images at a location corresponding to the subset of text regions;providing the area to a barcode decoder;receiving decoded data from the barcode decoder, the decoded data identifying a product;generating a price query containing the decoded data; andreceiving, responsive to the price query, the reference price.

9. The method of claim 7, wherein obtaining the reference price comprises: providing the image to a product recognition engine;receiving, from the product recognition engine, a location indicator within the image and a product identifier associated with the location;generating a price query containing the product identifier; andreceiving, responsive to the price query, a reference price.

10. A server for extracting price text from an image set, comprising: an image set processor configured to obtain, from an image sensor, input data comprising (i) a plurality of images depicting shelves supporting products, and (ii) for each of the images, a set of text regions and corresponding price text strings, wherein each text region is indicated by a set of pixel coordinates within the corresponding image obtained from the image sensor;the image set processor further configured to register the images to a common frame of reference;a selector configured to: identify a subset of the text regions having overlapping locations in the common frame of reference by retrieving from a repository a shelf edge location in the common frame of reference based on navigational data associated with each of the plurality of images and discarding any text regions that are not coincident with the shelf edge location;select one of the text regions from the subset; andpresent the price text string corresponding to the one of the text regions for comparison to a reference price.

11. The server of claim 10, the selector configured to identify the subset of text regions by identifying text regions having at least a predefined degree of overlap in the common frame of reference.

12. The server of claim 10, wherein the input data further comprises a confidence value corresponding to each of the price text strings.

13. The server of claim 12, the selector configured to select one of the text regions by selecting the one of the subset of text regions having the highest confidence value.

14. The server of claim 10, the selector further configured to: obtain a plurality of depth measurements corresponding to the plurality of images; andresponsive to obtaining the shelf edge location, perform a validation of the shelf edge location based on the depth measurements.

15. The server of claim 14, the selector configured to perform the shelf edge validation by: selecting a subset of the depth measurements falling within the shelf edge location obtained in the common frame of reference;clustering the subset of the depth measurements into a plurality of clusters;selecting a primary one of the clusters;fitting a plane to the primary cluster; anddetermining whether the plane has a preconfigured orientation.

16. The server of claim 10, further comprising a comparator configured to: obtain a reference price;compare the price text string corresponding to the one of the subset of text regions to the reference price; andpresent a result of the comparison.

17. The server of claim 16, further comprising an extractor configured to: extract an area of at least one of the images at a location corresponding to the subset of text regions;provide the area to a barcode decoder;receive decoded data from the barcode decoder, the decoded data identifying a product;the comparator configured to obtain the reference price by generating a price query containing the decoded data, and receiving, responsive to the price query, the reference price.

18. The server of claim 16, the comparator configured to obtain the reference price by: providing the image to a product recognition engine;receiving, from the product recognition engine, a location indicator within the image and a product identifier associated with the location;generating a price query containing the decoded data; andreceiving, responsive to the price query, a reference price.