Arrangement for broadband multispectral imaging

Abstract

Camera for broadband multispectral imaging, featuring - a broadband lens (1) for capturing light across an entire spectral range, - a first image acquisition unit (6) for acquiring light from a first sub-range of the entire spectral range located below a cutoff wavelength, - a second image acquisition unit (5) separate from the first image acquisition unit (6) for acquiring light from a second sub-range of the entire spectral range located above a cutoff wavelength, - a beam splitter (3) arranged between the lens (1) and the two image acquisition units (5, 6), with which the light of the entire spectral range can be split into light of the first part of the entire spectral range, which can be directed to the first image acquisition unit (6) by the beam splitter (3), and into light of the second part of the entire spectral range, which can be directed to the second image acquisition unit (5) by the beam splitter (3), - a common vessel (7) in which the two image acquisition units (5, 6) and the beam splitter (3) are housed, characterized by the fact that - the first image acquisition unit (6) is equipped with a first filter unit designed to divide the light of the first sub-area of ​​the entire spectral range into several spectral channels, and the second image acquisition unit (5) is equipped with a second filter unit designed to divide the light of the second sub-area of ​​the entire spectral range into several spectral channels, wherein all spectral channels of both parts are virtually superimposed during image acquisition.

Description

Field of invention

[0001] The present invention relates to the field of multispectral imaging. This is a field of image processing that has experienced strong growth in recent years. New image sensors, particularly for the SWIR range, have led to a significant expansion of applications. Both high spatial resolution and good spectral resolution are of interest. Spatial resolution is determined by the number of pixels, and spectral resolution by the number of detectable spectral channels. In the visible range (VIS), a large number of image sensors are available that, analogous to the human eye, provide three spectral channels: red (R), green (G), and blue (B) (RGB). These three channels, which ensure the spectral selectivity of the image sensors, are implemented in almost all image sensors by a mosaic-like filter structure, i.e., by chips with a mosaic filter on the chip (filter-on-chip) with the layout according to Bayer, 1976, US Patent No.3 971 065, “Color imaging array”, provided. The semiconductor base material used for these image sensors, which are typically designed as CCD / CMOS sensors, is silicon (Si).

[0002] In the visible spectrum, there is a vast number of applications that have permeated everyday life, in addition to numerous technical fields. Examples include smartphone use, production monitoring, food quality assessment, vehicle and aircraft operation, military applications in autonomous aircraft such as cruise missiles and drones, and the emerging field of artificial intelligence. It's important to note that the detection range of silicon-based image sensors, which extends to a maximum wavelength of 1100 nm, is not fully utilized in everyday applications (smartphone photography and video recording). To achieve a natural image appearance, the near-infrared (NIR) range is blocked by a filter. For technical applications, this blocking filter can be omitted, allowing silicon-based image sensors to capture the NIR range (700 nm to 1000 nm) as well.These sensors can already be used to solve a very wide variety of technical tasks.

[0003] However, there are also a number of tasks that require the detection of wavelength ranges outside the visible light (VIS) and near-infrared (N / A) spectrum. For numerous technical and biological detection applications, the near-infrared (N / A) range (700 nm to 1000 nm), adjacent to the long-wavelength end of the VIS range (400 nm to 700 nm), and the immediately following short-wavelength infrared (SWIR, 1000 nm to 1700 nm) range are of particular interest. This SWIR range is especially interesting because many technical, biological, and medical processes exhibit characteristic, image-like characteristics within it. Agriculture and forestry are a particular focus. Here, only inadequate results are achieved in the visible spectrum, as the vitality and biological properties of plants are primarily reflected in the SWIR range.Forestry is currently undergoing a transformation towards more climate-resilient trees. To assess growth in forest nurseries and in the forest, fast optical methods are needed in both the visible (VIS) and short-wavelength infrared (SWIR) ranges. Only a small portion of the SWIR range, extending to 1100 nm, can still be captured with silicon-based image sensors. For cost reasons, this limitation is utilized in a wide variety of sensor and camera designs. However, silicon-based image sensors fail when the detection range needs to be extended further towards longer wavelengths, i.e., to the range of 1100 nm to 1700 nm. For image acquisition in this range, the use of a different semiconductor base material, indium gallium arsenide (InGaAs), is essential. In contrast to the abundant base material silicon, the elements In, Ga, and As are available in significantly smaller quantities.InGaAs-based image sensors are therefore significantly more expensive than Si-based image sensors, but are indispensable for applications in the SWIR range.

[0004] To capture images across the entire broadband range (VIS / NIR / SWIR) from 400 nm to 1700 nm, two different types of image sensors based on different semiconductors are required: a silicon-based image sensor for the visible range and the adjacent NIR range up to 1000 nm (with the use of an IR cut filter, only for the visible range up to 700 nm) and an InGaAs-based image sensor for the SWIR range from 1000 nm to 1700 nm. In the NIR wavelength range, the spectral sensitivity curves of both image sensors overlap (see Fig. 1), so that this area can in principle be detected with both Si and InGaAs image sensors.

[0005] In both the VIS and NIR / SWIR ranges, high spectral selectivity, i.e., the division into an adequate number of spectral channels, is required to solve detection tasks.

[0006] It should be noted that different definitions for the NIR and SWIR spectral ranges are used in the literature. While the wavelength interval 400 nm to 700 nm is generally defined as the visible range, the following are common for NIR / SWIR: - NIR: 700 nm to 1000 nm, SWIR: 1000 nm to 1700 nm (in this application), - NIR: 700 nm to 1400 nm, SWIR: 1400 nm to 3000 nm (in the patent documents DE 10 2011 106 585 A1 and DE 10 2012 005 938 A1 discussed below).

[0007] However, these different definitions are not essential for the invention described in this application. State of the art

[0008] A characteristic feature of the current state of the art in multispectral imaging is the use of separate cameras for the visible (VIS) and black-and-white (SWIR) ranges. The reasons for this are manifold. Cameras for the SWIR range were developed later and were, and still are, used without a connection to the visible spectrum. If a connection is nevertheless necessary, two separate cameras are used. The requirements for the lenses differ and are therefore difficult to combine in a single lens. The same applies to the necessary light sources.

[0009] The approach of combining cameras with different spectral sensitivities in a single unit has been in use for a long time. In contrast to the relatively late problem of channel separation using mosaic filters (the retina of the human eye being a natural equivalent), which was only solved around 1980, two other fundamental principles have been in use for much longer: parallel channel separation by splitting the light path into multiple light paths (three-tube camera, three-chip camera) and sequential channel separation through the temporal sequence of filtering with different filters.

[0010] A technically complex arrangement of two different camera types, based on a silicon sensor (for the visible range) and an InGaAs sensor (for the near-infrared / swimmer's range), housed in a common container and using a shared lens, is described in patent documents DE 10 2011 106 585 A1 "Adaptation Lens" and DE 10 2012 005 938 A1 "Multispectral Zoom Lens and Camera System" of Carl Zeiss Optronics GmbH, Oberkochen, Germany. The focus is on the lens design, but the connection between the two camera types is also presented. The sensors of both cameras are spatially separated. However, they capture the same image via a semi-transparent mirror; that is, they receive light of different broadband wavelength ranges from the same section of an object, thus imaging the same section of the object.This type of connection between the image sensors of two cameras is referred to in this application as virtual overlay.

[0011] DE 10 2011 106 585 A1 describes in

[0012] the decomposition of a broad spectral range / wavelength range (400 nm to 3000 nm) into two sub-ranges directed to different sensors: into the visual range (400 nm to 700 nm) and the NIR range (700 nm to 1400 nm) or into the visual range (400 nm to 700 nm) and the SWIR range (1400 nm to 3000 nm). DE 10 2012 005 938 A1 describes in [0057-0058] the decomposition of a broad spectral / wavelength range (400 nm to 1700 nm) into three sub-ranges: Using a rotating beam splitter element, the visible range (400 nm to 700 nm), the NIR range (700 nm to 1400 nm), and the SWIR range (1400 nm to 1700 nm) are separated and directed to three sensors. These arrangements have the disadvantage that only broadband sensors are used, which only respond to light intensities within their respective wavelength ranges.The three sensors thus each capture the entire visual, NIR, and SWIR ranges, respectively. However, they are not designed for a finer spectral decomposition of their respective ranges. Multispectral acquisition by subdivision into individual channels capturing a narrow wavelength range, e.g., 50 nm to 100 nm, does not occur. Therefore, within their respective wavelength ranges, the sensors capture an image of objects that is independent of wavelength and determined only by light intensity, but they do not provide spectrally resolved information. This is due to the...

[0012] DE 10 2011 106 585 A1,

[0015] , is known to extract a very narrow wavelength range, in this case from 1530 nm to 1570 nm, from a broad wavelength range, there the NIR / SWIR range from 1000 nm to 3000 nm, using a single pivoted filter. However, no solution for the spectral resolution of the detected wavelength range from 1000 nm to 3000 nm is presented. Rather, the pivoted filter acts as a monochromator, which selects a wavelength range of only 40 nm (i.e., 2%) from the 2000 nm broad wavelength range for a specific laser application (there for detecting laser radiation with a wavelength of 1550 nm) and suppresses the two sections (1000 nm to 1530 nm and 1570 nm to 3000 nm, i.e., 98%). The information from these two sections is thus lost. DE 10 2011 106 585 A1 does not suggest using this information. Rather, it suggests that using this information, e.g.,Capturing and subjecting this information to spectral resolution is technically not feasible. Known solutions thus enable the capture of the shape of objects in the VIS, NIR, and SWIR ranges, e.g., the remote observation or detection of maritime objects (e.g., ships). However, due to the lack of spectral resolution, they do not allow for the determination of the material composition of objects or the vitality of plants or animals, as spectrally resolved information, particularly in the NIR and / or SWIR range, is required for this. This principle, proposed in the aforementioned patent documents, of the virtual superposition of the non-spectrally resolved images 31, 32 from two sensors to form a two-channel image 33 is illustrated in [reference missing]. Fig. 2. The images are composed of individual pixels P and are therefore spatially resolved, however, each pixel registers the entire spectral range 400 nm to 700 nm (VIS) or 1400 nm to 1700 nm (SWIR), which is not spectrally decomposed.

[0013] Patent application US 2021 / 0172795 A, USA 2021, "Methods and apparatus for imaging discrete wavelength bands using a mobile device," presents a filter-changing device for a smartphone, designed as a filter wheel. This would, in principle, eliminate the limitation to the three VIS wavelengths and RGB. However, this would require a smartphone without a Bayer layout, which is practically unavailable on the market. With a smartphone featuring a Bayer layout, which always includes an IR cut filter, the filter curves are superimposed and limited to the VIS range up to 700 nm. The detectable range on the silicon semiconductor is limited to 1100 nm, meaning that the VIS / NIR range extending to 1000 nm is easily detectable, while only an insufficiently small portion of the SWIR range, extending to 1700 nm, is detectable, up to 1100 nm.

[0014] Patent application CN114466122A, China 2022, “Multispectral camera device and method thereof”, presents a solution with three optical channels, implemented using splitting mirrors. Three separate cameras are employed. However, the significant effort required for setup, adjustment, and operation hinders its practical application.

[0015] In their conference paper, Feifan Lv, Yinqiang Zheng, Bohan Zhang, Feng Lu: “Turn a Silicon Camera Into an InGaAs Camera”, in: Proceedings of the 2019 IEEE / CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019, pp. 5987-5995, DOI: 10.1109 / CVPR.2019.00614, Feifan Lv, Yinqiang Zheng, Bohan Zhang, Feng Lu describe an arrangement with two cameras, one silicon-based and one InGaAs-based, which allows simultaneous image capture of an object with both cameras. For this purpose, a motor-driven filter wheel with one high-pass filter and five band-pass filters for the SWIR range is located between a lens and a beam splitter. This wheel allows multiple spectral channels to be selected sequentially, and images from both cameras can be captured in the selected channel.

[0016] The purpose of this arrangement, however, is not to enable broadband image acquisition from 400 nm to 1700 nm, but rather to enhance the silicon camera using a neural network and extend its detection range beyond the natural limit of 1100 nm in the infrared to 1200 nm. However, detection with a 1200 nm emitter is lacking and should, in principle, be impossible, as Jerry L. Hudgins explains in detail in "Wide and Narrow Bandgap Semiconductors for Power Electronics: A New Valuation," Journal of Electronic Materials, Vol. 32, No. 6, 2003, DOI: 10.1007 / s11664-003-0128-9, that the cutoff wavelength for silicon sensors cannot be exceeded at 1100 nm.

[0017] Furthermore, the arrangement described by Feifan Lv et al. has the following disadvantages: Due to the positioning of the filter wheel between the lens and the beam splitter, only a small portion of the incident broadband spectrum (VIS, NIR, SWIR) can be used for image acquisition—specifically, only the spectral range that passes through the currently active bandpass filter. Thus, only a fraction of the incident light intensity can be used for image capture, rendering the setup unsuitable as a broadband camera. Furthermore, the splitting of the light path immediately behind the filter wheel necessitates two separate lenses, each requiring separate refocusing for every scene change. Capturing a live image while the filter wheel is rotating is not possible.

[0018] Different spectral channels have different refractive indices, causing a shift in focus due to longitudinal aberration. This longitudinal aberration is not corrected, so a sharp image is only obtained in one spectral channel.

[0019] The setup is built on a solid base plate, but is nevertheless unprotected from external influences such as bumps or impacts. Furthermore, it is very bulky and heavy, and therefore unsuitable for mobile use. It can only be used under laboratory conditions.

[0020] The integration of image sensors from different spectral ranges in a single unit is currently achieved using the same semiconductor material (silicon) and represents the state of the art. Such cameras comprise a first silicon-based image sensor for the visible range (400 nm to 700 nm) and a second silicon-based image sensor whose detection range extends to 1100 nm, thus reaching just beyond the near-infrared (NIR) range (700 nm to 1000 nm). These spectrally resolved cameras are constructed using either a mosaic (Bayer layout) or parallel or sequential principles. The limit of these solutions, physically determined by the silicon band gap (1.12 eV), is 1100 nm. These silicon cameras are used in a variety of technical applications. They capture both the visible and the near-infrared (NIR) range, which is invisible to the human eye.This enables users to solve tasks that would not be solvable using only cameras for the visible area.

[0021] The particularly informative SWIR range above 1100 nm cannot be captured by such silicon cameras. It is only accessible through the addition of an InGaAs-based image sensor. Prior art arrangements feature two cameras (VIS / NIR camera and NIR / SWIR camera), each with its own lens and housing. However, these arrangements are impractical due to their large dimensions and complex, failure-prone design, as described, for example, in the aforementioned article by Feifan Lv et al. Fig. 6, shown, is not suitable for routine tasks under harsh conditions in industry (e.g. temperature stress) and agriculture and forestry (e.g. shock load during mobile use in the field).

[0022] There is therefore a high demand for a robust, compact and lightweight technical arrangement for multispectral broadband image acquisition across the entire VIS / NIR / SWIR detection range. Object of the invention

[0023] The object of the invention is therefore to provide such a robust, compact and lightweight technical arrangement for multispectral imaging across the entire VIS / NIR / SWIR detection range, i.e. in the wavelength range from 400 nm to 1700 nm. It is intended to optimally utilize the available radiation intensity across the entire detection range.

[0024] The arrangement is intended to be suitable for mobile use under harsh industrial conditions, e.g. in environments with high temperatures and high pressures, where it is intended to ensure the continuous acquisition of process parameters, e.g. temperature measurements, and material parameters, e.g. mechanical stresses in workpieces.

[0025] Furthermore, it should be usable in agriculture and forestry, e.g., for determining soil and forest condition. It should be suitable for mobile use in the field, carried by a human operator, by a mobile machine such as a tractor, agricultural machine, forestry machine (harvester), or a drone (flying at low altitudes up to 200 m), as well as for remote monitoring from the air (from altitudes above 200 m) or from space.

[0026] In particular, the arrangement, using a silicon-based image sensor (VIS sensor), should enable the acquisition of live images for orientation and object detection. Preferably, the arrangement should be equipped with cost-effective, commercially available standard components.

[0027] The layout should be quickly adaptable to different tasks specified by the user. Solution to the task

[0028] The object of the invention is solved by a camera for broadband multispectral image acquisition according to claim 1. Advantageous embodiments of the invention are disclosed in dependent claims 2 to 9. Brief description of the drawings Fig. Figure 1 shows the relative spectral sensitivity of two image acquisition units contained in the camera, the Si-based image sensor and the InGaAs-based image sensor, in relation to the spectral sensitivity of the human eye over the entire wavelength range to be captured. Fig. Figure 2 shows the virtual superposition of two different optical sensors Si, InGaAs, without further spectral subdivision, in accordance with the state of the art (DE 10 2011 106 585 A1 and DE 10 2012 005 938 A1). The result of the virtual superposition is an image signal whose individual pixels P each contain image information from a broad spectral range. Fig. These are the two spectral ranges VIS and SWIR. The two sensors register two spatially resolved images, but these do not receive any finer spectral resolution. Fig. Figure 3 shows the virtual superposition according to the invention of two different optical sensors housed in a common vessel, a Si sensor and an InGaAs sensor, with the spectral subdivision of the spectral range detected by the Si sensor by a mosaic filter and the spectral subdivision according to the invention of the spectral range detected by the InGaAs sensor, realized by a filter wheel. Fig. Figure 4 shows the virtual superposition according to the invention of two different optical sensors Si, InGaAs with the spectral subdivision of the spectral range detected by the Si sensor by a mosaic filter and the spectral subdivision of the spectral range detected by the InGaAs sensor also by a mosaic filter. Fig. Figure 5 shows a schematic representation of the basic structure of the camera with mosaic filter for the VIS range and filter wheel for the NIR / SWIR range. Fig. Figure 6 shows a schematic representation of the basic structure of the camera with two mosaic filters, one for the VIS range, the other for the NIR / SWIR range. Fig. Figure 7 shows the spectral selectivity achieved by the camera in the visible light range and the adjacent near-infrared and short-wave infrared ranges. Fig. Figure 8 shows a specific embodiment of the camera with mosaic sensor for the VIS range and filter wheel for the NIR / SWIR range. Fig. Figure 9 shows a specific embodiment of the camera with two mosaic sensors, one for the VIS range and another for the NIR / SWIR range. Fig. Figure 10 shows the spectrally different characteristics of phenomena (image of a filament and several spruce needles) when simultaneously recording 18 channels, 3 in the visible and 15 in the infrared range (NIR, SWIR). Wavelengths are given in nm. Fig. Figure 11 shows an overall view of the simple and quick filter change through a service opening provided for this purpose in the camera and the special tool used for this purpose. Fig. Figure 12 shows a cutaway drawing to illustrate the simple filter change through a service opening provided in the camera and the special tool used for this purpose. Detailed description of the solution

[0029] The camera according to the invention is described below with reference to the Fig. Sections 1 to 12 are described in detail. The following exemplary embodiment refers to the Fig. 5 and Fig. 8 a specific embodiment of the camera is specified.

[0030] The object of the invention is solved by a device with two filter units according to Fig. 3 (mosaic filter for the VIS range, filter changing device, preferably designed as a filter wheel, for the NIR / SWIR range) or Fig. 4 (mosaic filter for both the VIS and NIR / SWIR range) equipped camera for multispectral imaging with the in Fig. 5 or Fig. The basic structure is shown in Figure 6. The camera is equipped with two separate image acquisition units 5 and 5a, 6, each having a spectral channel subdivision realized by means of the filter units, which are housed in a common vessel 7 and in their combination ensure the acquisition of the entire spectral range from 400 nm to 1700 nm.

[0031] The first image acquisition unit 6 is designed to acquire at least a first sub-range of the entire spectral range, i.e., to acquire the spectral range from a wavelength of 400 nm up to an application-specific cutoff wavelength; it is preferably implemented as a silicon-based image sensor 6. The second image acquisition unit 5 is designed to acquire at least a second sub-range of the entire spectral range, i.e., to acquire the spectral range from this cutoff wavelength up to a wavelength of 1700 nm; it is preferably implemented as an InGaAs-based image sensor 5. The InGaAs-based image sensor 5 is rigidly connected to a precision linear displacement unit 2, which is preferably implemented as a piezoelectric linear actuator 2. The function of the linear displacement unit is explained below.

[0032] In reality, the detection ranges of the image sensors 5, 6 do not end abruptly (with a rectangular function) at the cutoff wavelength, but rather decrease gradually in a region surrounding the cutoff wavelength. As in Fig. As illustrated in Figure 1, the long-wavelength portion of the detection range of the first image acquisition unit 6 (Si-based image sensor, e.g., CCD / CMOS sensor) consequently overlaps with the short-wavelength portion of the detection range of the second image acquisition unit 5 (InGaAs-based image sensor). In principle, the cutoff wavelength can be set arbitrarily within the entire spectral range. For the present invention, a camera for multispectral image acquisition in the spectral range from 400 nm to 1700 nm, a cutoff wavelength in the transition region between the visible and infrared spectral ranges is selected. For technical reasons, it is set at 700 nm.

[0033] The camera features a broadband lens 1 that is transparent across the entire spectral range from 400 nm to 1700 nm and is therefore capable of coupling incident light across the entire spectral range (VIS / NIR / SWIR) into the camera. The lens 1 can be integrated into a wall of the vessel 7 ( Fig. 5, Fig. 6) or be attached to a wall of the vessel 7 ( Fig. 8) The lens 1 defines a light path, i.e., a common beam path for incident light of the entire spectral range, within the camera. A beam splitter 3 is arranged behind the lens 1 in this common beam path. It transmits and reflects specific portions of the light incident through the lens. This divides the common beam path into a first image acquisition channel for the reflected light portion and a second image acquisition channel for the transmitted light portion. The first image acquisition unit 6 is positioned in the first image acquisition channel, and the second image acquisition unit 5 is positioned in the second image acquisition channel. The two image acquisition channels and the associated image acquisition units 5 and 6 are thus combined in a single technical unit.

[0034] The beam splitter 3 directs the reflected light component to the first image acquisition unit 6, while the transmitted light component is directed to the second image acquisition unit 5. Preferably, the surface of the beam splitter 3 exposed to the incident light is inclined at an angle of 45° relative to the common beam path, so that the reflected light component is deflected at an angle of 90° perpendicular to the transmitted light component. This allows for a simple camera design with a cuboid or cube-shaped container 7 in which the two image acquisition units 5, 6 are positioned on two mutually perpendicular walls of the container 7. However, other angles of inclination can also be chosen. They should be in the range of 30° to 60°. The two image acquisition units 5, 6 are always positioned on those wall regions toward which the two image acquisition channels are directed.

[0035] In its simplest form, the beam splitter 3 is a simple, non-wavelength-selective beam splitter, which can be designed as a semi-transparent mirror. Largely independent of the wavelength, such a beam splitter reflects a certain proportion (relative light intensity x) of the light incident through the broadband lens 1 (relative light intensity 1) and transmits the remaining proportion (relative light intensity 1 - x). The transmitted (transmitted) light is slightly redirected parallel to itself by refraction upon entering and exiting the semi-transparent mirror 3, thus reaching the second image acquisition unit 5, positioned in the second image acquisition channel, without any change in direction. Preferably, a beam splitter 3 is chosen which, when inclined at an angle of 45° relative to the common beam path, splits the incident light equally between the two image acquisition channels (x = 0.5).Depending on the user's requirements, a beam splitter 3 can also be selected, which ensures a different percentage division, where x can take any value between 0 and 1.

[0036] Such a simple beam splitter 3 is very cost-effective, but has the disadvantage that it supplies both image acquisition units 5, 6 with light from the entire spectral range, even though each unit can only capture a portion of the entire spectral range (below or above the cutoff wavelength). As a result, approximately 50% of the available light intensity is lost unused. It cannot be used for image acquisition or other measurement purposes. The unusable portion of the light intensity even leads to an undesirable heat load on the image acquisition units 5, 6.

[0037] To overcome these disadvantages, the beam splitter is preferably designed as a dichroic mirror 3, which splits the incident light across the entire spectral range (400 nm to 1700 nm) into a short-wavelength component below the cutoff wavelength, which is reflected by the dichroic mirror 3 and fed to the first image acquisition unit, and a long-wavelength component above the cutoff wavelength, which is transmitted by the dichroic mirror 3 and fed to the second image acquisition unit. The dichroic mirror 3 is thus designed as a long-pass filter. The dichroic mirror 3 can also be designed as a short-pass filter, which transmits light with a wavelength below the cutoff wavelength but reflects light with a wavelength above the cutoff wavelength. In this short-pass configuration, the positions of the two image acquisition units 5 and 6 must be interchanged.

[0038] Dichroic mirrors 3 with different cutoff wavelengths are available. For the present application, a camera for multispectral imaging in the spectral range from 400 nm to 1700 nm, the cutoff wavelength of the dichroic mirror 3 should be adapted to the detection ranges of the two image acquisition units 5, 6. Dichroic mirrors 3 with cutoff wavelengths in the interval from 700 nm to 1100 nm are therefore preferred. A dichroic mirror with a cutoff wavelength of 700 nm is particularly preferred. Preferably, the dichroic mirror 3 is inclined at an angle of 45° relative to the common beam path, so that it deflects the light with a wavelength below the cutoff wavelength into the first image acquisition channel, which is inclined at an angle of 90° relative to the common beam path, and directs it onto the first image acquisition unit positioned in the first image acquisition channel.Other tilt and deflection angles and a correspondingly adjusted position of the first image acquisition unit are of course possible.

[0039] The common vessel 7 protects all components installed within it from mechanical stresses.

[0040] A preferred embodiment of the camera, comprising a cost-effective, commercially available Si-based image sensor with an IR blocking filter, will now be explained in more detail.

[0041] The camera has the basic structure described above. Its first image acquisition unit is equipped with a visual image sensor, such as those used in standard smartphones, industrial color cameras, or digital photography. These image sensors are commercially available in a wide variety of designs. They are based on the semiconductor material silicon. As an optically active material, silicon allows the capture of the electromagnetic spectrum in the range of 400 nm to 1100 nm, i.e., the VIS / NIR wavelength range. To capture images with such silicon-based image sensors that convey a natural image impression, these sensors are always equipped with an infrared (IR) blocking filter. This IR blocking filter can be positioned in front of the sensor as an additional optical component, for example, in wildlife and surveillance cameras.If daylight is available, the IR blocking filter is retracted into the beam path, thus suppressing IR light. At night, the IR blocking filter is retracted from the beam path, allowing IR light to be detected by the silicon sensor. In most applications, however, the IR blocking filter is permanently attached to the sensor. The IR blocking filter ensures that only the visible range from 400 nm to 700 nm is used, while the IR range is suppressed, resulting in images comparable to those of the human eye. Spectral selectivity, i.e., the division into three spectral channels—red, green, and blue (RGB)—is achieved in the vast majority of technical applications, among other methods, by a mosaic-like filter structure, which corresponds to the layout described in the prior art section according to Bayer, US Patent No. 3,971,065.This technique is also referred to as "filter-on-chip," a term already used several times above, and generally employs wavelength-selective filters. However, other filter types, such as polarizing filters, can also be used. The silicon-based image sensor positioned in the first image acquisition channel is preferably implemented as a CCD / CMOS sensor with a Bayer mosaic layout (mosaic filter sensor). It allows the acquisition of live images comparable to those of the human eye.

[0042] The second image acquisition unit, which ensures the recording of the NIR / SWIR wavelength range from 700 nm to 1700 nm, is based on the semiconductor material InGaAs. The short-wavelength limit of the spectral sensitivity of such InGaAs-based image sensors was, as Fig. Figure 1 illustrates that until 2022, the bandwidth was approximately 800 nm, which was insufficient for measurements in the 700 nm to 800 nm range. However, in 2023, InGaAs-based image sensors came onto the market that are not limited to the SWIR range but also cover the NIR range and even the entire VIS range. Specifically, the SenSWlR™ image sensors IMX990 and IMX991 from SONY™ should be mentioned here. The qualitative development of the spectral sensitivity of these newer InGaAs-based image sensors is shown in Figure 1. Fig. Figure 1 is reproduced. The exact experimental data from SONY™ are available in: “Features of image sensors with SenSWIR technology”, Sony Semiconductor Solutions Corporation (2023), https: / / www.sonysemicon.com / files / 62 / flyer_industry / IMX990_991_992_993_Flyer_en.pdf, p. 2. These newer InGaAs-based image sensors thus enable measurements down to the required wavelength of 700 nm, i.e., measurements in the detection range of 700 nm to 1700 nm. The second image acquisition unit will therefore be equipped with such a newer InGaAs-based sensor, which covers the entire NIR / SWIR wavelength range from 700 nm to 1700 nm. The spectral selectivity according to the invention is achieved in the NIR / SWIR wavelength range by a filter changing device (filter wheel 4), preferably with a total of fifteen channels, as shown in Fig. 5 shown. Alternatively, as in Fig. Figure 6 illustrates that spectral selectivity in the NIR / SWIR range can also be achieved by an NIR / SWIR image sensor 5a equipped with a mosaic filter (filter-on-chip) with a 3 x 3 structure for nine channels or a 4 x 4 structure for 16 channels. These two possibilities for the technical realization of spectral selectivity in the NIR / SWIR wavelength range are explained in more detail below. Thus, the camera divides the entire spectral range from 400 nm to 1700 nm into spectrally effective channels.

[0043] The camera can be equipped with a cost-effective, simple beam splitter. However, to avoid its aforementioned disadvantages, a wavelength-sensitive beam splitter, preferably designed as a dichroic mirror, is preferably used in the camera. Its cutoff wavelength is selected where the detection ranges of the two image sensors touch or overlap, in this case at 700 nm. The dichroic mirror directs the visible (VIS) component (400 nm to 700 nm) of the incident broadband light by reflection onto the silicon-based image sensor in the first image acquisition channel and the near-infrared (NIR) / swift-infrared (SWIR) component (700 nm to 1700 nm) of the incident broadband light by transmission onto the InGaAs-based image sensor in the second image acquisition channel. Both image sensors are thus only exposed to light of the wavelength range to which they are sensitive.The incident broadband light can therefore (in an idealized view) be utilized at 100% when using a dichroic mirror, while 50% of the light intensity is lost when using a simple, non-wavelength-sensitive beam splitter. In reality, the proportion of utilized light remains below 100%, since the sensitivities of the image sensors, as well as the characteristics of the IR cut filter and the dichroic mirror, are not defined by sharp boundaries (rectangular functions), but rather decrease or increase in a wavelength band surrounding the cutoff wavelength. This camera allows the combined acquisition of live images with a natural image appearance in the visible range using the silicon-based image sensor and images in the NIR / SWIR range using the InGaAs-based image sensor.

[0044] The spectral selectivity required for the camera in the NIR / SWIR range can be achieved in different ways: First possibility: In the second image acquisition channel, a filter changing device, preferably a rotatable filter wheel 4 that can be fitted with filters, is arranged between the dichroic mirror 3 and the InGaAs-based image sensor 5. The filter changing device comprises a freely selectable number of spectrally effective optical filters, i.e., bandpass filters, which define adjacent spectral channels in the required wavelength range (700 nm to 1700 nm). They are installed in windows of the filter wheel. The NIR / SWIR range is thus divided into a freely selectable number of spectral channels. The filter changing device achieves the necessary parameters of spectral selectivity and operating speed. Preferably, a division into fifteen spectral channels is used. Changing the filters during operation of the filter wheel 4 requires optical compensation of the optical path length.To achieve this, the InGaAs-based image sensor is rigidly connected to a precision linear displacement unit 2, which is preferably designed as a piezoelectric linear actuator. The linear displacement unit 2 allows a linear displacement of the InGaAs-based image sensor 5 in the direction of the optical axis, i.e., along the second image acquisition channel. The displacement required for each filter to compensate for the light path length is determined once by a series of focus adjustments.

[0045] Subsequently, in regular operation, the respective displacement values ​​are retrieved and the InGaAs-based image sensor 5 is moved to the focus point for each filter by the piezo linear drive 2.

[0046] To expand or adapt the application range, quick and easy replacement of the individual filters 41 in the filter wheel 4 is desirable. This is made possible by a service opening 43 in the housing, which allows easy access to the individual filters 41. A special tool 42, preferably a vacuum gripper 42, is used for handling, i.e., gripping and positioning the filters. This easy replacement of the individual filters 41 ensures rapid adaptation to changing measurement tasks in the NIR / SWIR range. In addition, this also makes it possible to switch the detection range to other spectral ranges, such as the UV range (UV - ultraviolet) with wavelengths below 400 nm.

[0047] Second option: Instead of using a filter changing device, the InGaAs-based image sensor 5a is implemented as a mosaic filter sensor, which ensures the division of the NIR / SWIR range into spectral channels. This mosaic filter sensor has the advantage that no filter wheel and no optical compensation of the light path length are required, as illustrated in Fig. 6 (schematic representation) and Fig. 9 (specific embodiment).

[0048] Fig. Figure 7 shows the spectral selectivity achieved by the camera with the two image sensors 5, 6 and filter wheel 4 in a common vessel 7 across the entire VIS / NIR / SWIR spectral range from 400 nm to 1700 nm. This arrangement provides three channels in the visible (VIS) range and fifteen channels in the adjacent NIR / SWIR range. The channels in the NIR / SWIR range (700 nm to 1700 nm) are assigned spectrally effective optical filters that select narrowband adjacent frequency bands (full width at half maximum approximately 67 nm) of the NIR / SWIR light spectrum. This results in fine channel separation in the NIR / SWIR range. The arrangement offers the crucial advantage of always having a live image available thanks to the mosaic filter sensor in the VIS range.

[0049] The camera uses the in the Fig. 3 and Fig. 4 schematically represented virtual superposition according to the invention with simultaneous multispectral image acquisition, which is explained below.

[0050] At the in Fig. Figure 3 schematically depicts the virtual superposition of images 34, 35 from two different optical sensors housed in a common vessel: a silicon (Si) and an InGaAs (InGaAs) sensor. Eighteen channels are generated for each pixel of the captured image 36: three channels (RGB) in the visible (VIS) range, realized using a mosaic filter (filter-on-chip), and fifteen channels in the near-infrared (NIR) / swift-infrared (SWIR) range, realized using a filter wheel. According to the Bayer layout, each pixel in the image from the silicon sensor comprises one sector for red (R), one sector for blue (B), and two sectors for green (G), thus ensuring spectral resolution in the VIS range. To ensure spectral resolution in the NIR / SWIR range, 15 filters are sequentially introduced into the beam path of the InGaAs sensor by the rotating filter wheel. The spectral range captured by the two sensors is thus spectrally resolved by decomposing it into eighteen channels that are virtually superimposed.

[0051] The in Fig. 4. The schematically represented virtual superposition of the images from two different optical sensors, the image of a Si sensor 37 and the image of an InGaAs sensor 38, differs from the one in Fig. Figure 3 illustrates this by using a mosaic filter for spectral subdivision in the VIS range, as before, while using a mosaic filter for spectral subdivision in the NIR / SWIR range. Instead of a rotating filter wheel, the InGaAs sensor is implemented as an InGaAs sensor with a 3x3 mosaic filter. In Figures 37 and 38 of the Si sensor and the InGaAs sensor, one pixel is highlighted in each case. The pixel sizes must be, as Fig. Figure 4 shows that the pixels do not match. Different pixel sizes are corrected during image processing using suitable algorithms. In the resulting image 39, the virtual superposition creates twelve virtually superimposed channels for each pixel: three channels (RGB) in the VIS range using mosaic filters (filter-on-chip), and nine channels in the NIR / SWIR range, also using mosaic filters (filter-on-chip). The central wavelengths of the bandpass filters of the nine channels do not need to be equidistant, as illustrated by Figure 38, which shows the selected central wavelengths (in nm) of the channels for a pixel. The range between 1130 nm and 1220 nm is resolved more finely than the range between 1220 nm and 1660 nm. The user can therefore select the central wavelengths of the filters for specific applications. The same applies to the bandwidths of the filters. The mosaic filter of the InGaAs sensor can be further varied, e.g.,by implementing it as a 4x4 mosaic filter (with 16 channels in the NIR / SWIR range). However, the only state of the art is the virtual superposition of non-spectrally resolved images, as in [reference missing]. Fig. 2 schematically represented, known. The ones in the Fig. 3 and Fig. A schematically represented virtual superposition of spectrally resolved images, in which all spectral channels are virtually superimposed, is not known from the prior art. Alternative embodiments and further advantageous configurations of the invention

[0052] As mentioned above, the dichroic mirror 3 can also be configured as a shortpass filter. In this case, the image acquisition channels for the NIR / SWIR range and for the VIS / NIR range are reversed. The positions of the image sensors 5, 6 and the filter wheel 4 must be adjusted accordingly.

[0053] The number of channels in the NIR / SWIR range can be varied as desired, with an individual channel width selectable for each channel. This allows the spectral selectivity in the NIR / SWIR range to be optimally adapted to the task to be solved by the user.

[0054] It is also possible to exclude selected sections of the entire spectral range from detection. For example, if the visible (VIS) and swir (SWIR) ranges are of interest, but not the near-infrared (NIR) range (700 nm to 1000 nm), the NIR range can be excluded by using a suitable filter.

[0055] In an advantageous embodiment of the invention, the filter changing device comprises, in addition to the bandpass filters for wavelength separation, i.e., the spectrally selective filters, other optically active elements, e.g., polarizing filters. This is achieved, for example, by having a greater number of filter positions in the filter changing device, e.g., the windows in the filter wheel 4, than the number of spectrally selective bandpass filters. The excess filter positions are equipped with other optically active elements, such as polarizing filters for determining the degree and angle of polarization. With a device equipped in this way, additional tasks in the mechanical stress analysis of workpieces or in the functional layer analysis can be performed.

[0056] Besides evaluating radiation distribution for phenotypic detection and measurement tasks, measuring thermal radiation using InGaAs-based image sensors (SWIR sensors) is of interest. InGaAs-based sensors achieve a lower temperature limit of approximately 550 K for high-temperature applications. A combination of RGB imaging and temperature overlay is also conceivable and important for many applications.

[0057] In summary, the following parameters can be analyzed: color, wavelength composition (depending on filter wheel configuration), polarization degree / angle, temperature (from approx. 550 K) and depth measurement by evaluating the focus position, depending on the chosen lens.

[0058] The principle of the camera according to the invention can be transferred to other wavelength ranges, e.g. the ultraviolet (UV) range. For example, a UV / VIS camera can be realized in this way.

[0059] The invention is used in technical fields such as rapid material analysis, as well as in the fields of biology, food science, agriculture and forestry, and medicine. Reference symbol list 1 lens 2 Linear shift unit for focus adjustment of the InGaAs-based image sensor 3 beam splitters, dichroic mirror 4 filter wheel 5 InGaAs-based, non-wavelength-sensitive image sensor for the NIR / SWIR range 5a InGaAs-based mosaic image sensor for the NIR / SWIR range 6 Si-based mosaic image sensor (VIS sensor with three channels red, green, blue (RGB)) 7 common vessel 20 Light path defined by the lens (common beam path) 20a First image acquisition channel (400 nm to 700 nm) 20b Second image acquisition channel (700 nm to 1700 nm) 31 VIS sensors, 1 spectral channel for the entire VIS range 32 SWIR sensors, 1 spectral channel for the entire SWIR range 33 Virtual overlay, 2 spectral channels for the entire VIS and SWIR range 34 VIS sensor, filter-on-chip, 3 spectral channels (RGB) in the VIS range 35 SWIR filter wheel sensor, 15 spectral channels in the IR / SWIR range (SWIR 1 - 15) 36 virtual overlays, 18 spectral channels across the entire VIS / IR / SWIR range 37 VIS mosaic sensor, filter-on-chip, 3 spectral channels (RGB) in the VIS range 38 SWIR mosaic, filter-on-chip, 9 spectral channels in the SWIR range (SWIR 1 - 9) 39 virtual overlays, 12 spectral channels across the entire VIS / IR / SWIR range 41 filters 42 Special tools 43 Service opening FN spruce needles GW filament M engines P pixels Example of implementation

[0060] Fig. 5 (schematic representation) and Fig. Figure 8 (specific embodiment) shows an exemplary embodiment of the camera setup for broadband multispectral imaging. Light from the entire spectral range (400 nm to 1700 nm) passes through the broadband lens 1 along the common beam path 20 into the interior of the vessel 7. The vessel has the dimensions (width B, height H, depth T) B x H x D = 110 mm x 96 mm x 143 mm. The common beam path 20 is divided by a dichroic mirror 3, arranged behind the lens 1 and configured as a long-pass filter, into a first image acquisition channel 20a for light from the VIS range (400 nm to 700 nm) and a second image acquisition channel 20b for light from the NIR / SWIR range (700 nm to 1700 nm).

[0061] The first image acquisition channel 20a directs the light from the visible range onto a silicon-based image sensor 6, which is designed as a Bayer mosaic filter sensor with an IR cut filter for the visible range. The light is permanently applied to this image sensor 6 and generates a continuously available RGB color image for the visible range. This silicon-based image sensor 6 enables object detection, camera alignment, and signal evaluation in three channels—red, green, and blue (RGB)—within the visible range.

[0062] The second image acquisition channel 20b first directs the light from the NIR / SWIR range onto a filter wheel 4 with 15 windows, in which 15 filters are installed, dividing the wavelength range from 700 nm to 1700 nm equidistantly. The filters are moved through image acquisition channel 20b by the rotational movement of the filter wheel 4. The light transmitted by the filters reaches the InGaAs-based image sensor 5. A newer InGaAs sensor (SONY IMX990) is used, whose detection range extends into the visible spectrum, thus completely covering the NIR / SWIR range. The movement of the filter wheel 4 is stepwise and synchronous with the frame rate of the InGaAs-based image sensor 5, caused by the filter wheel drive. The filters used are bandpass filters based on optical glass. During filter changes, e.g.,Variations in the thickness of the filter glass can lead to a shortening or lengthening of the optical path length and thus to blurring. To avoid this, the InGaAs-based image sensor 5 is rigidly connected to a linear displacement unit 2, which allows it to be moved along the optical axis, i.e., along the second image acquisition channel 20b. The focus point is determined once for each filter by means of the linear displacement unit 2 as it passes through the filter wheel 4. This focus point is then approached for each filter by the precision linear displacement unit 2 during subsequent operation of the filter wheel 4. The light transmitted by the filters reaches the InGaAs-based image sensor 5. The linear displacement unit 2 and the filter wheel 4 are driven by two motors M, which are controlled by a motor controller.The motor control unit is connected to camera electronics, which in turn are connected to a computer used to program the motor control unit. The camera electronics receive the images from the two image sensors 5 and 6 and forward them to the camera output.

[0063] At the camera output, the image from the three-channel silicon-based image sensor 6 (VIS sensor) and the image from the fifteen-channel InGaAs-based image sensor 5 (NIR / SWIR sensor) are provided and transmitted to a computer. This results in a total of eighteen images per capture, ranging from 400 nm to 1700 nm, available for further analysis. It is worth noting that despite the sequential filter changes in the NIR / SWIR range, a live image in the VIS range is always available. This enables the rapid location of the objects under investigation, which is virtually indispensable for analyzing the vitality of plants in the wild.

[0064] In Fig. Figure 10 shows an image captured with the multispectral broadband camera according to the invention, comprising 18 channels. It includes the three channels of the VIS range (RGB), a broadband (non-spectrally resolved) channel for the wavelength range of 900 nm to 1650 nm to obtain an overview image in the NIR / SWIR range, and 11 narrowband channels whose central wavelengths extend at equidistant intervals of 50 nm over the wavelength range of 950 nm to 1600 nm. The bandwidth of the filters (FWHM - Full Width at Half Maximum) is set at 50 nm.

[0065] The image in Fig. Figure 10 shows typical effects such as the visualization of a faintly luminous filament of an incandescent lamp (GW), which is not visible in the visible range and is clearly visible in the images at 1500 nm, 1550 nm, and 1600 nm. In the image at 1600 nm, the position of the filament (GW) is additionally marked by an arrow pointing towards it.

[0066] Similarly, the moisture content, and thus the vitality, of spruce needles can be determined by their moisture level. In the image at 1000 nm, two groups of spruce needles (FN) are visible, depicted with almost identical contrast, again indicated by arrows pointing to the two groups. In the image at 1450 nm, however, the right group of spruce needles is depicted with weaker contrast than the left group. This indicates that the spruce needles in the right group are drier and therefore have lower vitality than the spruce needles in the left group.

[0067] In Fig. Figure 11 demonstrates the possibility of a quick and easy filter change for the version of the camera with filter changing device (filter wheel 4).

[0068] For this purpose, the camera has an additional service opening 43 in the camera housing 7, which is accessible both in the standby and operating state. For replacement, the individual filters 41 are each gripped with a special tool 42, preferably a vacuum gripper 42, and removed from the filter wheel 4 or placed in the filter wheel 4. Fig. Figure 12 illustrates the arrangement of filter wheel 4, filter 41, special tool 42 and service opening 43 in a sectional view. QUOTES INCLUDED IN THE DESCRIPTION

[0000] This list of documents cited by the applicant was automatically generated and is included solely for the reader's convenience. The list is not part of the German patent or utility model application. The DPMA accepts no liability for any errors or omissions. Cited patent literature

[0000] US 3 971 065 [0001, 0041] DE 10 2011 106 585 A1 [0006, 0010, 0011, 0012, 0028] DE 10 2012 005 938 A1 [0006, 0010, 0011, 0028] US 2021 / 0172795 A

[0013] CN 114466122A

[0014] Zitierte Nicht-Patentliteratur

[0000] USA 2021, „Methods and apparatus for imaging discrete wavelength bands using a mobile device

[0013] Feifan Lv, Yinqiang Zheng, Bohan Zhang, Feng Lu: „Turn a Silicon Camera Into an InGaAs Camera“, in: Proceedings of the 2019 IEEE / CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019, pp. 5987-5995, DOI: 10.1109 / CVPR.2019.00614

[0015] Jerry L. Hudgins: „Wide and Narrow Bandgap Semiconductors for Power Electronics: A New Valuation“, Journal of Electronic Materials, Vol. 32, No. 6, 2003, DOI: 10.1007 / s11664-003-0128-9

[0016] Features of image sensors with SenSWIR technology“, Sony Semiconductor Solutions Corporation (2023), https: / / www.sonysemicon.com / files / 62 / flyer_industry / IMX990_991_992_993_Flyer_en.pdf, S. 2

[0042]

Claims

[1] Camera for broadband multispectral imaging, comprising - a broadband lens (1) for capturing light across an entire spectral range, - a first image acquisition unit (6) for acquiring light from a first sub-range of the entire spectral range located below a cutoff wavelength, - a second image acquisition unit (5) separate from the first image acquisition unit (6) for acquiring light from a second sub-range of the entire spectral range located above a cutoff wavelength, - a beam splitter (3) arranged between the lens (1) and the two image acquisition units (5, 6), with which the light of the entire spectral range can be split into light of the first part of the entire spectral range, which can be directed to the first image acquisition unit (6) by the beam splitter (3), and into light of the second part of the entire spectral range, which can be directed to the second image acquisition unit (5) by the beam splitter (3), - a common vessel (7) in which the two image acquisition units (5, 6) and the beam splitter (3) are housed, characterized by , that - the first image acquisition unit (6) is equipped with a first filter unit designed to divide the light of the first sub-area of ​​the entire spectral range into several spectral channels, and the second image acquisition unit (5) is equipped with a second filter unit designed to divide the light of the second sub-area of ​​the entire spectral range into several spectral channels, wherein all spectral channels of both parts are virtually superimposed during image acquisition. [2] Camera according to claim 1, characterized by , that the entire spectral range extends over a wavelength range from 400 nm to 1700 nm, wherein the first sub-range of the entire spectral range covers the wavelength range of visible light from 400 nm to 700 nm and the second sub-range of the entire spectral range covers the wavelength range of near / shortwave infrared light from 700 nm to 1700 nm. [3] Camera according to claim 1 or 2, characterized by , that the beam splitter (3) is designed as a wavelength-selective beam splitter in the form of a dichroic mirror (3) with which the light of the entire spectral range can be split into light of the first sub-range, which can be directed by the dichroic mirror (3) to the first image acquisition unit (6), and into light of the second sub-range, which can be directed by the dichroic mirror (3) to the second image acquisition unit (5). [4] Camera according to claim 1, 2 or 3, characterized by , that the cutoff wavelength, which separates the first sub-range from the second sub-range, is set to 700 nm. [5] Camera according to claim 2, 3 or 4, characterized by, that the first image acquisition unit is a Si-based image sensor (6) with an integrated filter device, which is designed as a mosaic filter sensor with IR blocking filter, with which the spectral range of visible light can be decomposed into the three channels, red, green and blue, and that the second image acquisition unit is an InGaAs-based image sensor (5), the filter unit of which is designed as a filter changing device (4), preferably as a filter wheel (4), comprising optical filters with which the spectral range of near / short-wave infrared light can be decomposed into several channels, preferably into 15 channels. [6] Camera according to claim 2, 3 or 4, characterized by, that the first image acquisition unit is a Si-based image sensor (6) with an integrated filter unit designed as a mosaic filter sensor with IR blocking filter, with which the spectral range of visible light can be decomposed into the three channels, red, green and blue, and that the second image acquisition unit is an InGaAs-based image sensor (5a) with an integrated filter unit designed as a mosaic filter, with which the spectral range of near / short-wave infrared light can be decomposed into several channels. [7] Camera according to claim 5, characterized by , that the filter changing device (4) of the InGaAs-based image sensor (5) has polarization filters in addition to the optical filters. [8] Camera according to claim 5 or 7, characterized by, that the InGaAs-based image sensor (5) is rigidly connected to a linear displacement unit (2), designed as a piezo linear drive, so that by moving the InGaAs-based image sensor (5) it is possible to compensate for different light path lengths caused by the filters or to shift the focus for depth measurement. [9] Camera according to claim 5, 7 or 8, characterized by , that the common vessel (7) has a service opening (43) so that it is possible to replace the filters (41) in the filter changing device (4) using a special tool (42), preferably a vacuum gripper (42), supplied through the service opening (43).

Images