Method related to the operation of pattern recognition devices, in particular load scales

The present invention relates to a method covering the operation of separate images-knowing devices for detecting objects with a respective one of the image recognition underlying object data set.

From European Patent <patcit id="pcit0001" dnum="EP0443419B1"><text>EP 0443419 B1</text></patcit> are image-recognizing self-service scales and equipped without Wägefunktionalität terminals for object recognition are known which are used in the retail trade and make it possible to identify goods via video cameras. To this end, recently captured images stored in an image memory merchandise images are compared by means of a respective video camera of the goods. The stored goods pictures form a goods data record, which is placed the image recognition based.

Usually, in a first start of the self-service scales and the terminal a standard goods data record is stored. During operation, the stored standard goods record then can a learning process ( "teaching") are expanded to recognize goods may, for which no goods images were previously available. Alternatively, images of goods can be imported to the newly added products from a library or the like. Have pictures to goods no longer needed can be deleted or overwritten.

From the European patent application <patcit id="pcit0002" dnum="EP2199779A1"><text>EP 2199779 A1</text></patcit> it is known to store in a terahertz scanner for a goods identification required comparative images on a server, each time when a product is to be recognized, a server access is required.

The present invention has for its object to provide an opportunity to increase the image recognition rate image-knowing devices.

This object is achieved by a method having the features of independent claim 1, and in particular concerning a method to operate the underlying networked with a central location separate images-knowing devices for detecting objects with a respective one of the image recognition object data set, wherein the object data sets of the equipment on a network can be centrally managed.

In particular, with the devices to devices of a retail company, which are each equipped with an image pickup device to a goods detection, as explained above perform, can. In particular, the objects are goods. In the image pickup device, it is preferably at least one camera, in particular video camera. The devices are preferably to scales, especially retail scales and / or self-service scales, information terminals and / or other terminals for goods identification. The data maintenance of the individual object data sets may - in particular - carried out by the central office. In particular, objects to be recognized by the respective associated device object data in the form of object images are included in an object data set. In particular, in the respective image recognition underlying object data sets of equipment to local property records. The central administration of the object data sets of the devices can be matched.

The image recognition is especially the ease of operation at a scale or an information terminal. A customer can get over the image-recognizing device additional information on a product by bringing the goods into the detection range of the image-forming device characteristic. Then, the image recognition of the appliance can be automatically or manually run started. The device recognizes the goods, looking in a goods database for information associated with the product, and displays this information on a display.

In particular, it may be in the image forming device featuring a retail scale, identifying the goods to be weighed on image recognition and the basic price of the goods from a price database, preferably an ERP system, automatically reads. The image recognition of the product allows a user of the scale for easy operation by the scale automatically identify goods and a manual entry of a PLU ( "Price Lookup Code") may be omitted.

Particularly in the case of image recognition the images captured by the video camera are identified by the object data. The object data can be determined based on comparison images, for example by extracting from an image-related features or patterns and these are processed with an appropriate data-reducing method or filtering process and stored as object data. By comparing the object data with corresponding processed data of a currently recorded product or goods image automatic identification of products or goods can take place.

It is advantageous if the image-recognizing unit converts locally available object data or change or override. This change or adjustment can be made in a continuous learning process during the image recognition. For example, if an automatic image recognition fails and there is a manual correction of the automatic selection, the image-recognizing device may change its local object data so that the data quality is improved. For example, on a display a selection of possible matches, and an operator manually selects the correct results out. Based on the data of a proper mapping the image-recognizing device can now link and charge data for the current product image with the object data. For example, influences of changes in illumination conditions can be compensated. It creates a continuous quality improvement process that improves with an increasing number of product image detections automatic image recognition rate.

It can also be provided that an automatic adjustment of the local object data from the central point variety, may in particular disabled or enabled, are. The central body can enable or disable the automatic adjustment of the local object data in particular for a single image featuring utensil or for a group of pictures featuring end devices. For example, the detection rate of a device is close to the optimum, automatic adjustment can be blocked, for example, to receive an increased processing speed or to use, for example for quality assurance on multiple devices always uniform object data.

Image detection rates of the devices can be recorded especially for a particular object of the central body and preferably also evaluated. Devices with high error rates in the image recognition can thus be recognized and targeted maintenance. The image recognition rate can be documented for each device.

Additionally or alternatively, in particular, can be exchanged between the devices and / or objects to be recognized associated between the central point and the devices object data over the network and / or the central body.

In particular, in a device produced for example by teaching, a particular object associated with object data - especially in the event that with this object data high image recognition rate is connected - to the central body and / or - in particular on the central point - at least another device are received. It can be made available to all other devices available so the object data of the device with the highest detection rates.

Preferably or in at least one optionally distributed object database of the network, in particular the central location, associated objects to be recognized object data stored. The respective unit can then the respective object data set - in particular at a first start - or part thereof - in particular in the context of an update or an update - are received by the at least one object database. When updating particular object data are received, provide the high or maximum image recognition rates.

Preferably, the respective, in particular conveyed property data set to all or only selected objects assigned in at least one object database stored object data. Specifically selected objects assigned object data is not stored in the at least one object database. The at least one object database can be linked to a database or to lock object data specific objects.

Not selected objects associated object data can not be transmitted to the respective device, are deleted from the respective object data record and / or disabled in each object data set. In an object data set located deactivated object data then are no longer part of the object record for the purposes of the present invention. Disabled object data can for instance be provided with a flag. This data, for example, seasonal items or not, or is no longer in the program is imported articles can be selectively blocked. A smaller number of object data in an object data set resulting in the image recognition to a higher processing speed and - since at least not currently in range contained items can not be recognized incorrectly - to lower error rates.

After training the invention, the object data sets of the equipment is regularly backed up by the central office and / or in the network. After failures of a device in which the respective object data set is lost, the last saved version of each object data set can then at least be entered again. This allows to operate the device with the present to the past time of backup Wägedatensatz on. The defect may be for example, a hard drive crash. The time intervals between the times at which the object data records are saved may vary.

According to another embodiment of the invention are or the devices are classified and are jointly classified devices same object data - in particular from the at least one object database - transmitted. In particular, the at least one object database for at least a portion of the objects to be detected in each case contains, ie for each object, different from each other, differently classified object data, and sent to the respective device according to classified data object. The object data can therefore still classification data, eg sausage department / cheese department, branch, national company and / or lighting situation be saved. Several similar devices or devices with similar classification data, as lighting, product range and location, can be combined to a group of devices and with the same or the classification of the corresponding object data are supplied. This allows the installation costs are reduced and the recognition rate can be improved.

By the present invention, the detection rate can be significantly increased by the image recognition. Basically, the invention can be generally used for optical control device using image recognition, especially in the production or quality assurance.