Method for the identification of stored information

Abstract

The invention relates to a method for the identification of templates, in the form of stored information. An identifying suggestion is automatically generated for the template. The generation of said suggestion can occur separately from the actual identification process, for example when the method for generation of the suggestion is time-consuming. A generation of the suggestion during the identification process is also possible, for example, when the method is in the position to generate the suggestion in real-time or almost in real-time. The identification suggestions for the observer, thus, are rapidly generated, permitting a representative object recognition.

Description

[0001] The manual characterization of samples by people is a time-consuming and expensive process. In order to speed up this process, methods are used which generate a characterization proposal. The processing employee must either accept the proposal or modify it if the proposal is not correct. Characterization methods are used in particular in classification methods in which a set of parameters for identifying objects is determined using training examples. The training data record must representatively cover all the peripheral conditions of the identification task here. In order to identify street scenes, samples of several thousand to tens of thousands of manually processed images containing all the potentially occurring classes of objects, for example passenger cars, trucks, two-wheeled vehicles, pedestrians, weather conditions etc., are required.

[0002] From the processing of colored video images, a method is known in which an object, for example a tree, is characterized manually in the first image in an image sequence (J. R. Ohm, P. Ma: “Feature-Based Cluster Segmentation of Image Sequences”, Int. Conf. on Image Processing, Vol. III, 1997, pages 178-181). With this predefined information an attempt is made to identify again the same object in the next image of the image sequence. In this context, the color information of the object is preferably used to distinguish the object from the rest of the image. This method is not suitable for gray value images. The manual processing of the object is very time-consuming during the characterization process.

[0003] A method is known (A. K. Jain, Yu Zhong, S. Lakshamanan, “Object matching using deformable templates”, IEEE Trans. on Pattern Analysis and Machine Intelligence, Vol. 18, No. 3, March 1996, p. 267-278) in which a database of automatically generated contour patterns is used to characterize objects (patterns) in image data. The contour patterns are automatically adapted to the objects and superimposed on them. The image data together with the contour patterns superimposed on them are then presented to a viewer on a screen.

[0004] The object of the invention is to find a novel method for automatically characterizing patterns which ensures that the pattern and characterization coincide to an optimum degree.

[0005] The invention is described in claim 1. Advantageous refinements and developments can be found in the subclaims.

[0006] The invention has the advantage that a large gain in time in the characterization process is achieved during the automatic generation of characterization proposals, even when there are faulty proposals owing to the mass of the samples to be characterized. It is also advantageous that automatic methods are used for generating the characterization proposals and said methods can be very time-consuming and demanding in terms of computing as the characterization proposals can be generated independently of the characterization process. The characterization proposals which are generated are stored in a database until they are checked by the processing employee and are presented to the processing employee while the characterization process is taking place.

[0007] When automatic methods are used for generating characterization proposals, which methods process the proposals very quickly and present them to the processing employee, it is advantageous that the proposals can be generated directly during the characterization process.

[0008] The invention is explained in more detail by means of an exemplary embodiment.

[0009] The pattern which is to be characterized is present in a computer system in electronic form. The pattern is recorded, for example, with a video camera or a document is scanned into a computer. Acoustic patterns, which are present in electronic form, are also characterized.

[0010] A characterization proposal is generated for the pattern. This proposal can be generated separately from the actual characterization process, for example if the method for generating the proposal is very time-consuming. However, it is also possible to generate the proposal during the characterization process, for example if the method is capable of generating the proposals in real time or virtually in real time.

[0011] The pattern and the generated characterization proposal are presented to the processing employee in suitable form on a display by means of a computer program. In the case of images, for example, the original image is presented together with the characterization proposal which is presented in the form of a closed line configuration, for example. The line configuration is superimposed on the original image so that the processing employee can see both the line configuration and the original image. If the object to be characterized is characterized with sufficient precision by the generated proposal, the proposal is accepted by the processing employee and stored in the system as a characterization. If the proposal is too imprecise, there are a number of possibilities. On the one hand, the processing employee will delete the incorrect proposal (or modify it) himself and generate his own characterization proposal manually. On the other hand, it is possible for the proposal to be rejected by the processing employee. The proposal is then used as a basis for a further method in order to still arrive at a correct characterization proposal. Furthermore, it is possible to delete the proposal completely and to try to generate a suitable proposal using a more complex method.

[0012] In order to cope with a large number of samples, the characterization method is carried out using the Internet. Throughout the world, the processing and, if appropriate, collection of samples is carried out by means of loosely organized groups of between several hundred to a thousand persons.

[0013] In the context of the use of the invention within a data network, a device according to the invention for carrying out the method is to be advantageously considered as a system which is divided in two, composed of a server system and a processing device.

[0014] In this context, the server system has a device for storing patterns, a unit for generating characterization proposals for the stored patterns, a memory for storing the characterizations which are assigned to the patterns and a communications device for communicating with a processing device for evaluating and processing the characterization proposals. As counterpart, the processing device for evaluating and processing characterization proposals for the patterns is configured in such a way that it comprises a display for displaying the patterns and the associated characterization proposals and an input unit for inputting the assessment of the characterization of the pattern and/or for deleting or modifying the characterization. Furthermore, the processing device has a communications unit for communicating with a server system for storing patterns and characterizations and for generating characterization proposals.

[0015] With the objective of allowing as many characterization processes as possible to run simultaneously, it is advantageous if the server system is connected to a plurality of processing devices.

Claims

1. A method for characterizing patterns which are present as stored information, characterization proposals being automatically generated in a computer system, and the pattern and the characterization proposal being presented to a processing employee by means of a computer program, characterized in that the processing employee can evaluate and/or modify the characterization proposal which is assigned to the pattern.

2. The method as claimed in claim 1, characterized in that the characterization proposals which are generated are stored in a database until they are checked by the processing employee, and are presented to the processing employee on a screen while the pattern is being characterized.

3. The method as claimed in claim 1, characterized in that the characterization proposal is generated directly on the screen while the pattern is being characterized.

4. The method as claimed in one of the preceding claims, characterized in that the pattern is an image and the image is presented on the display together with the characterization proposal.

5. The method as claimed in claim 4, characterized in that the characterization proposal is presented in the form of a closed line configuration, and in that the line configuration is superimposed on the image.

6. The method as claimed in one of the preceding claims, characterized in that the proposal is accepted by the processing employee and stored in the system as a characterization.

7. The method as claimed in one of the preceding claims 1 to 5, characterized in that the proposal is deleted or modified by the processing employee, and in that a characterization proposal is generated manually.

8. The method as claimed in one of the preceding claims 1 to 5, characterized in that the proposal is rejected by the processing employee, and in that the proposal is used as a basis for a further method.

9. The method as claimed in one of the preceding claims 1 to 5, characterized in that the proposal is deleted completely, and in that a suitable proposal is generated using a more complex method.

10. The method as claimed in one of the preceding claims, characterized in that for a large number of samples the characterization method is carried out using the Internet.

11. A server system for carrying out the method as claimed in one of claims 1 to 10, characterized in that the server system has a device for storing patterns, a unit for generating characterization proposals for the stored patterns, a memory for storing the characterizations which are assigned to the patterns, and a communications device for communicating with a processing device for evaluating and processing the characterization proposals.

12. A processing device for evaluating and processing characterization proposals for patterns within the scope of the method as claimed in one of claims 1 to 10, characterized in that the processing device has a display for displaying the patterns and the associated characterization proposals, an input unit for inputting the assessment of the characterization of the pattern and/or for deleting or modifying the characterization and a communications unit for communicating with a server system for storing patterns and characterizations, and for generating characterization proposals.