US-12619095-B2 - Production of cameras with reduced rejection rate

US12619095B2US 12619095 B2US12619095 B2US 12619095B2US-12619095-B2

Abstract

A method for producing a camera. The method includes: providing prefabricated components; adjusting at least two of these prefabricated components relative to one another in accordance with at least one specified optimality criterion; and adhesively bonding the components to one another in the adjusted state; wherein prior data characterizing a specific specimen of at least one of the prefabricated components, and/or measured data in respect of the optical performance of the combination of the components adjusted with respect to one another, are mapped by a trained machine learning model onto a prediction for the optical performance that the camera will deliver once it has run through at least one additional production step after the adhesive bonding; and this prediction is used as feedback for an influencing action on the production process.

Inventors

Jens Dorfmueller

Assignees

ROBERT BOSCH GMBH

Dates

Publication Date: 20260505
Application Date: 20220119
Priority Date: 20210310

Claims (11)

1 . A method for producing a camera, comprising the following steps: providing prefabricated components; adjusting at least two of the prefabricated components relative to one another in accordance with at least one specified optimality criterion; and adhesively bonding the at least two of the prefabricated components to one another in the adjusted state; wherein: prior data characterizing a specific specimen of at least one of the prefabricated components, and/or measured data in respect of an optical performance of the combination of the at least two of the prefabricated components adjusted relative to one another, are mapped by a trained machine learning model onto a prediction for an optical performance that the camera will deliver once the camera has run through at least one additional production step after the adhesive bonding; and the prediction is used as feedback for an influencing action on the production process, wherein the prediction takes into account changes in the optical performance of the camera between the adhesive bonding and final completion.
2 . The method according to claim 1 , wherein: several candidate specimens are provided for at least one of the at least two of the prefabricated components; using prior data characterizing each candidate specimen of the candidate specimens, the machine learning model ascertains a respective prediction for an optical performance of a camera which contains the candidate specimen; and a combination of the candidate specimens for which the prediction satisfies a specified criterion, is selected for further production of the camera.
3 . The method according to claim 1 , wherein: during the adjustment, the machine learning model ascertains multiple times, based on measured data relating to optical performance of a combination of the at least two of the prefabricated components in a current spatial arrangement relative to each other, a prediction for the optical performance of the camera which results when the at least two of the prefabricated components are adhesively bonded to one another in this arrangement; and in response to the prediction satisfying a specified criterion the at least two of the prefabricated components are adhesively bonded to one another.
4 . The method according to claim 3 , wherein, during the adjustment, optimization with respect to the prediction provided by the machine learning model is given priority over optimization with respect to the specified optimality criterion.
5 . The method according to claim 1 , wherein, in response to the prediction for the optical performance of the camera satisfying a specified criterion, the production process is terminated.
6 . The method according to claim 1 , wherein the prior data characterize: a modulation transfer function (MTF) of an optical component, and/or measurement results from a quality test of a component in the context of prefabrication, and/or a supplier of a component, and/or at least one tool used for the production of a component.
7 . The method according to claim 1 , wherein the measured data characterize: a modulation transfer function of a combination of the at least two of the prefabricated components adjusted relative to one another, and/or dimensions of a spatial arrangement of the at least two of the prefabricated components adjusted to one another.
8 . The method according to claim 1 , wherein the prediction for the optical performance characterizes a modulation transfer function of the camera as completed.
9 . A non-transitory machine-readable data medium on which is stored a computer program including machine-readable instructions for producing a camera, the instructions, when executed by one or more computers, cause the one or more computers in combination of a production facility for cameras controller by the one or more computers, to perform the following steps: providing prefabricated components; adjusting at least two of the prefabricated components relative to one another in accordance with at least one specified optimality criterion; and adhesively bonding the at least two of the prefabricated components to one another in the adjusted state; wherein: prior data characterizing a specific specimen of at least one of the prefabricated components, and/or measured data in respect of an optical performance of the combination of the at least two of the prefabricated components adjusted relative to one another, are mapped by a trained machine learning model onto a prediction for an optical performance that the camera will deliver once the camera has run through at least one additional production step after the adhesive bonding; and the prediction is used as feedback for an influencing action on the production process, wherein the prediction takes into account changes in the optical performance of the camera between the adhesive bonding and final completion.
10 . One or more computers comprising: a non-transitory machine-readable data medium on which is stored a computer program including machine-readable instructions for producing a camera, the instructions, when executed by the one or more computers, cause the one or more computers in combination of a production facility for cameras controller by the one or more computers, to perform the following steps: providing prefabricated components; adjusting at least two of the prefabricated components relative to one another in accordance with at least one specified optimality criterion; and adhesively bonding the at least two of the prefabricated components to one another in the adjusted state; wherein: prior data characterizing a specific specimen of at least one of the prefabricated components, and/or measured data in respect of an optical performance of the combination of the at least two of the prefabricated components adjusted relative to one another, are mapped by a trained machine learning model onto a prediction for an optical performance that the camera will deliver once the camera has run through at least one additional production step after the adhesive bonding; and the prediction is used as feedback for an influencing action on the production process, wherein the prediction takes into account changes in the optical performance of the camera between the adhesive bonding and final completion.
11 . A production facility for cameras, the production facility configured to: provide prefabricated components; adjust at least two of the prefabricated components relative to one another in accordance with at least one specified optimality criterion; and adhesively bond the at least two of the prefabricated components to one another in the adjusted state; wherein: prior data characterizing a specific specimen of at least one of the prefabricated components, and/or measured data in respect of an optical performance of the combination of the at least two of the prefabricated components adjusted relative to one another, are mapped by a trained machine learning model onto a prediction for an optical performance that the camera will deliver once the camera has run through at least one additional production step after the adhesive bonding; and the prediction is used as feedback for an influencing action on the production process, wherein the prediction takes into account changes in the optical performance of the camera between the adhesive bonding and final completion.

Description

FIELD The present invention relates to the production of cameras made of prefabricated components. BACKGROUND INFORMATION The optical performance of a camera depends on whether the actual configuration of the beam path from the scenery being imaged to the sensor being used for image capture corresponds to the advance planning for this purpose. The optical components used, e.g., lenses and objectives, must therefore conform to their respective specifications and be mounted in the correct spatial arrangement relative to one another. The optical performance of the camera ultimately produced results from an interaction between the precision of the optical components on the one hand and the precision of the spatial arrangement of these components on the other hand. When the optical components are produced with very tight tolerances, the accuracy of pick-and-place production techniques is sufficient for obtaining cameras with acceptable optical performance. If lower-cost optical components produced with wider tolerances are used, the resulting deviations of the beam path from the advance planning can be compensated for at least in part by adjusting the components during production. In this adjustment (also referred to as “active alignment”), components are moved relative to one another according to one or more degrees of freedom, and a configuration in which the optical performance is good is fixed by adhesive bonding of the components. German Patent Application No. DE 10 2014 220 519 A1 describes an exemplary method for such an adjustment. SUMMARY The present invention provides a method for producing a camera. For example, this camera can in particular be designed to monitor the environment of a vehicle, or at least a portion thereof. For example, such cameras can enable a driver of the vehicle to see into otherwise difficult-to-see areas of the vehicle environment, e.g., behind the vehicle or in a “blind spot” not visible via mirrors. However, driving assistance systems or systems for the at least partially automated driving of a vehicle can also utilize camera images from the vehicle environment. According to an example embodiment of the present invention, the method begins with providing prefabricated components. For example, these components can in particular include optical components, e.g., lenses, objectives, and image sensors. However, the components can also include, e.g., electronic circuitry for a further evaluation and transmission of the images captured by the image sensor, as well as a housing. At least two of the prefabricated components are adjusted relative to one another in accordance with at least one specified optimality criterion. In the adjusted state, i.e., when the specified optimality criterion is satisfied, then the components are adhesively bonded to one another. This adhesive bonding can be initiated, e.g., by activating a light curing adhesive using UV light. The optimality criterion can, e.g., include that an optical performance of the combination of the components being adjusted relative to one another and measured in accordance with any desired metric lies above a specified threshold. In addition, according to an example embodiment of the present invention, prior data that characterize a specific specimen of at least one of the prefabricated components, and/or measured data, are used with respect to the optical performance of the combination of the components adjusted relative to one another. It is thus possible to use only prior data, only measured data (e.g., from an “active alignment” machine), or any combination of prior data and measured data. These data are mapped by a trained machine learning model onto a prediction for the optical performance that the camera will deliver after it has run through at least one additional production step after the adhesive bonding. This prediction is used as feedback for an action of influencing the production process. If measured data with respect to the optical performance are obtained during the adjustment of the prefabricated components, then this already delivers an indication of how good the ultimate optical performance of the camera will be. However, it has been found that the state achieved in the context of the adjustment can yet change by way of the at least one further processing step which the camera runs through after the adhesive bonding of the components. For example, if another electronic circuit board is installed in a housing in which the optical components are already located, the housing may be shaken and/or heated. Furthermore, the adhesive used is not yet fully cured immediately after the adhesive bonding of the components. The components are indeed already protected from the adjustment achieved then becoming completely lost due to shaking or other handling of the camera, but the adhesive can still shrink during the course of the final curing, which can take several more minutes. The spatial arrangement of the adhesively