Bionic compound eye imaging device

Abstract

The invention belongs to the technical field of imaging, and relates to a bionic compound eye imaging device which comprises a head cover and imaging equipment arranged on the head cover. The head cover is of a semispherical structure, the interior of the semispherical structure is hollow, and a shell space is formed so that image processing equipment can be arranged; the image equipment comprises a central unit arranged at the vertex of the head cover and a plurality of edge units arranged around the central unit at intervals in an array manner, so as to form a bionic compound eye structure; the center unit and the edge unit are both connected with the image processing device so as to realize target detection imaging; the number of the edge units is eight, and the edge units form an annular array in the circumferential direction of the head cover. According to the invention, large-view-field imaging and multi-view-angle time sequence image super-resolution reconstruction can be realized.

Description

Technical Field

[0001] The present application relates to the field of imaging technology, and in particular to a bionic compound eye imaging device. Background Art

[0002] In recent years, low-altitude unmanned platforms such as multi-rotor and fixed-wing drones have developed rapidly. In order to meet the needs of aerial surveillance, target detection and identification, they must have good target search, identification and tracking capabilities. In order to improve the situational awareness and confrontation capabilities of low-altitude unmanned platforms in complex environments, large field of view and high resolution are essential for optical imaging systems. At the same time, miniaturization and lightweight are also indispensable to enhance the load capacity of unmanned platforms.

[0003] In order to improve the optical imaging system, in the past two decades, researchers have designed many types of optical bionic compound eyes by imitating the imaging mechanism of biological compound eyes. These eyes have multiple imaging apertures, each of which is called a sub-eye. They can obtain sub-images from different perspectives and provide information input for visual applications. Since the structure of parallel compound eyes is easier to implement than that of overlapping compound eyes, most bionic compound eyes at this stage are designed with reference to parallel compound eyes, and each sub-eye has an independent imaging optical path.

[0004] Existing bionic compound eye technologies mainly include: (1) TOMBO planar bionic compound eye imaging system: Each small unit can image the collected image separately, and by analyzing the subtle differences between the images collected by each unit, a high-resolution image can be synthesized. However, due to the limitations of the structure itself, the resolution of the imaging unit is low, the image distortion is serious, and the reconstructed image resolution has no significant advantage over the single-aperture optical system.

[0005] (2) APCO imaging system: It includes multiple parallel independent imaging units. The microlenses are formed by photopolymers and directly attached to the glass layer. A metal layer is plated under the glass layer. Microholes are distributed at corresponding positions on the metal layer to form a microhole array, so that the imaging units do not interfere with each other. This design reduces the thickness of the imaging system, but the image resolution is still low, the edge imaging quality is poor, and there is obvious distortion.

[0006] (3) Curved compound eye camera SCECam with optical image transfer system: This camera consists of a curved spherical shell, 117 sub-eyes, an optical relay system, and a CMOS image sensor. The radius of the spherical shell is 20 mm, the diameter of each sub-eye is 3 mm, and the focal length is 5 mm. The optical relay system composed of 7 lenses corrects the focal surface of the spherical shell to a focal plane to match the image sensor, and the field of view reaches 122.4°. However, the achieved field of view still has room for improvement.

[0007] (4) Bionic compound eye large field of view micro camera for small drones: The total focal length of the optical system is 4 mm, the F number is 4, the field of view angle can reach 106°, and the sub-eye uses a double-cemented lens combination to reduce system aberrations. The resolution can reach 0.5 m at a flight altitude of 500 m. However, the achieved field of view angle is still small. Summary of the invention

[0008] Based on this, it is necessary to provide a bionic compound eye imaging device to address the above technical problems, which can achieve large field of view imaging and multi-view time-series image super-resolution reconstruction.

[0009] A bionic compound eye imaging device, comprising: a head cover and an imaging device arranged on the head cover; The head cover is a hemispherical structure, the interior of the hemispherical structure is hollow, forming a shell space for arranging an image processing device; The imaging device comprises: a central unit arranged at the top of the hood and a plurality of edge units arranged in an interval array around the central unit to form a bionic compound eye structure; the central unit and the edge units are both connected to the image processing device to achieve target detection imaging.

[0010] In one embodiment, the number of the edge units is eight, and they form a circular array along the circumference of the hood.

[0011] In one embodiment, the central unit and the edge unit each include: a lens module and a base; The lens module corresponds to the base one by one, and the lens module is arranged on the head cover by corresponding to the base.

[0012] In one embodiment, the lens module adopts a three-piece structure.

[0013] In one embodiment, the optical path of the lens module is perpendicular to the section corresponding to the head cover position where the base is located.

[0014] In one embodiment, it also includes: a tail cover and a heat dissipation device; The tail cover is a circular ring structure, the diameter of which is equal to the diameter of the hemispherical structure, so that one end of the tail cover is connected to the edge of the head cover; The heat dissipation device is arranged in the inner space of the tail cover.

[0015] In one embodiment, the target detection distance is: ; In the formula, is the pixel size of the imaging device, is the number of pixels occupied by the detected target, is the focal length of the imaging device, For the actual height, is the actual object distance.

[0016] In one embodiment, the image processing device includes: an image stitching component and an image super-resolution component, which are used for image stitching and image super-resolution reconstruction respectively.

[0017] In one embodiment, the image stitching component includes: an image preprocessing module, an image registration module, and an image fusion module connected in sequence; The image preprocessing module is used to reconstruct the acquired original image by using a histogram specification method and a histogram equalization method to obtain a preprocessed image; The image registration module is used to use a feature-based registration method to perform feature extraction, feature matching and perspective transformation matrix calculation on the pre-processed image to obtain a registered image; The image fusion module is used to fuse the registered images using a weighted average fusion algorithm to obtain a stitched image.

[0018] In one embodiment, the image super-resolution component includes connected: a generator model and a discriminator model; The generator model is used to reconstruct the spliced ​​image to obtain a reconstructed image; The discriminator model is used to determine whether the original image or the reconstructed image belongs to the generated image, and obtain the final imaging of the target.

[0019] The above-mentioned bionic compound eye imaging device is designed for low-altitude unmanned platform scene perception tasks (such as aerial surveillance, target detection and recognition, etc.), and a large-field-of-view compact bionic compound eye imaging device is designed. The camera is used to divide the field of view into several areas to achieve seamless division of the predetermined field of view, and the camera arrangement rules are confirmed according to the mapping relationship between the field of view and the camera to achieve the desired field of view. It can meet the performance requirements of an effective field of view of 160° and a detection distance of 2km for typical ground vehicle targets. It has the advantages of large field of view, small distortion and high resolution, and has the post-processing functions of multi-channel image display, image stitching, and image super-resolution. It realizes large-field-of-view imaging and multi-view time-series image super-resolution reconstruction, improves the effective field of view and detection distance of target detection and information capture, improves imaging resolution and image quality, and supports low-altitude visual algorithm research. In addition, this application uses the spherical coordinate system of the spherical shell as the design reference and optical reference to make the structure more compact, and the array camera group is modularly designed to realize the synchronous image acquisition function of the multi-camera module, providing technical support for the low-cost, lightweight and intelligent development of the optoelectronic sensor system of the low-altitude unmanned platform. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1A side view of the appearance of a bionic compound eye imaging device in one embodiment; Figure 2 is an internal side view of a bionic compound eye imaging device in one embodiment; Figure 3 A software architecture diagram of an image processing device of a bionic compound eye imaging device in an embodiment; Figure 4 It is a flowchart of the operation of the image processing device software of a bionic compound eye imaging device in one embodiment; Figure 5 A splicing flow chart of a bionic compound eye imaging device in an embodiment; Figure 6 A schematic diagram of the distribution of imaging devices of a bionic compound eye imaging device in one embodiment; Figure 7 A schematic diagram of the dimensions of a bionic compound eye imaging device in an embodiment; Figure 8 A schematic diagram of a central unit optical system of a bionic compound eye imaging device in an embodiment; Fig. 9 is an MTF curve diagram of a central unit optical system of a bionic compound eye imaging device at a cutoff frequency in an embodiment; Fig.10 A point arrangement diagram of a central unit optical system of a bionic compound eye imaging device in one embodiment; Fig.11 A schematic diagram of an edge unit optical system of a bionic compound eye imaging device in an embodiment; Fig.12 A schematic diagram of the overall structure of a module of a bionic compound eye imaging device in an embodiment; Fig.13 This is one of the schematic diagrams of module dimensions of a bionic compound eye imaging device in one embodiment; Fig.14 This is a second schematic diagram of module dimensions of a bionic compound eye imaging device in an embodiment; Fig.15 A top-down result diagram of a simulation of the field of view of a camera of a bionic compound eye imaging device in an embodiment; Fig.16 A side view result diagram of a camera field of view stitching simulation of a bionic compound eye imaging device in an embodiment; Fig.17 A stitched panoramic image of a bionic compound eye imaging device in an embodiment; Fig.18 This is a super-resolution image result of a bionic compound eye imaging device in an embodiment.

[0021] Reference numerals: Hood 1; Imaging device 2, central unit 21, edge unit 22, lens module 23, base 24; Image processing device 3; Heat dissipation device 4; Rear end cover 5. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solutions and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application.

[0023] It should be noted that all directional indications in the embodiments of the present application (such as up, down, left, right, front, back, etc.) are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0024] In addition, the descriptions of "first", "second", etc. in this application are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of this application, "multiple groups" means at least two groups, such as two groups, three groups, etc., unless otherwise clearly and specifically defined.

[0025] In this application, unless otherwise clearly specified and limited, the terms "connection", "fixation", etc. should be understood in a broad sense. For example, "fixation" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection, an electrical connection, a physical connection, or a wireless communication connection; it can be a direct connection, or an indirect connection through an intermediate medium, or it can be the internal connection of two elements or the interaction relationship between two elements, unless otherwise clearly defined. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0026] In addition, the technical solutions between the various embodiments of the present application can be combined with each other, but it must be based on the fact that ordinary technicians in the field can implement it. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by this application.

[0027] The present application provides a bionic compound eye imaging device, such as Figure 1 to Figure 2 As shown, in one embodiment, it includes: a head cover and an imaging device.

[0028] The head cover is a hemispherical structure, and the interior of the hemispherical structure is hollow to form a shell space for arranging the image processing equipment.

[0029] The imaging device is arranged on the hood, and includes: a central unit and a plurality of edge units; the central unit is arranged at the top of the hood, and a plurality of edge units are arranged in an array around the central unit to form a bionic compound eye structure; the central unit and the edge units are both connected to the image processing device (for example, through a USB interface) to image the central area of ​​the target through the central unit, and to image the edge area of ​​the target through the edge units, thereby realizing target detection imaging. Specifically, the central unit and the edge units can both be cameras.

[0030] Preferably, the number of the edge units is eight, and they form a circular array along the circumference of the hood to increase the detection field of view, so that the field of view of the imaging device reaches 160°.

[0031] Further preferably, the central unit and the edge unit both include: a lens module and a base; the lens module and the base correspond one to one, and the lens module is arranged on the head cover through the corresponding base.

[0032] More preferably, the lens module adopts a three-piece structure to cooperate with the optical design, optimize the focal length and field of view, balance the geometric aberration, and improve the imaging quality.

[0033] More preferably, the optical path of the lens module is perpendicular to the section corresponding to the head cover position where the base is located.

[0034] In one embodiment, it also includes: a tail cover and a heat dissipation device.

[0035] The tail cover is a circular ring structure, and the diameter of the circular ring structure is equal to the diameter of the hemispherical structure, so that one end of the tail cover is connected to the edge of the head cover to form a whole.

[0036] The heat dissipation device is arranged in the inner space of the tail cover.

[0037] In one embodiment, the image processing device includes: an image stitching component and an image super-resolution component. The image stitching component is used for image stitching, and the image super-resolution component is used for super-resolution reconstruction of the region of interest of the stitched image. The specific method of selecting the region of interest is an existing technology and will not be repeated here.

[0038] like Figure 3 The software architecture of the image processing device shown can simultaneously capture and display the center field camera image and the edge field camera image, and display the frame rate of each image in real time.

[0039] like Figure 4 The operation flow chart of the image processing device software shown is as follows: the software uses MFC to create an interface for displaying images for interaction, and displays the frame rate in real time; prepares kernel resources for the center field of view camera and the edge field of view camera, such as creating threads and frame locks; issues control instructions to open the center field of view camera and the edge field of view camera; obtains the center field of view camera image and the edge field of view camera image, and displays them on the interface; destroys all kernel resources and releases all camera resources when closing the software.

[0040] 1. Image stitching component The image stitching component includes: an image preprocessing module, an image registration module and an image fusion module, and the image preprocessing module, the image registration module and the image fusion module are connected in sequence.

[0041] The image preprocessing module is used to reconstruct the acquired original image by adopting the histogram regulation method and the histogram equalization method to obtain the preprocessed image; specifically: the histogram regulation method and the histogram equalization method are adopted to change the size, scale and shape of the acquired image to adapt to different stitching requirements, and the image enhancement technology is used to improve the brightness, contrast and color saturation of the image to improve the imaging effect of the final stitching to obtain the preprocessed image.

[0042] The image registration module is used to use a feature-based registration method to perform feature extraction, feature matching and perspective transformation matrix calculation on the preprocessed image to obtain a registered image. The basic idea is to find the overlapping parts between the images to be stitched, and determine the correspondence between the overlapping parts, and perform appropriate transformations on the images based on this correspondence; specifically: use the SIFT scale-invariant feature detection algorithm to extract feature points from the preprocessed image; use the nearest neighbor Hamming distance ratio algorithm to filter and match the extracted feature points; use the RANSAC random sampling consensus algorithm to iteratively calculate the perspective transformation matrix of the image to obtain a registered image.

[0043] The image fusion module is used to fuse the registered images using a weighted average fusion algorithm to obtain a stitched image. The stitched image is a relatively smooth and seamless stitching result, which solves the possible color difference problem and ghosting problem in the overlapping area between the images to be stitched.

[0044] like Figure 5The stitching flow chart shown in the figure inputs 9 images for preprocessing, then uses the SIFT algorithm to detect image feature points and calculate feature vectors, then uses the BFMatcher algorithm for feature matching, and filters out bad matches by using the ratio test method; after preliminarily filtering out bad matches between some feature points by using the ratio test, uses the RANSAC-based robust algorithm to calculate the optimal single mapping transformation matrix between the remaining two sets of feature points, and performs perspective transformation on the image according to the transformation matrix; finally, uses the weighted average fusion method to complete the image stitching and output the stitching result.

[0045] 2. Image super-resolution component The image super-resolution component uses an image super-resolution algorithm based on SRGAN to select an image area of ​​fixed size at any position in a spliced ​​large-format, large-scene image for super-resolution reconstruction at 2x and 4x scales, providing high-quality image support for subsequent target detection and recognition tasks, including: a generator model and a discriminator model, and the generator model is connected to the discriminator model.

[0046] The generator model is used to reconstruct the spliced ​​image to obtain the reconstructed image; specifically: the generator model adopts a structure called SRResNet, but when calculating the loss function, it needs to use the intercepted VGG19 model for calculation. Here, VGG19 is only used as a network for calculating image features, and it itself is not a generator module, that is, VGG19 is already trained and does not update its parameters (gradients are not updated), but only participates in calculating image features. For the structure of the generator model, the generation network first extracts features through a convolution layer, and then adds multiple residual blocks, each of which contains two convolution kernels: A convolutional layer is used, followed by a batch normalization layer (BN layer) and PReLU as the activation function. A series of residual blocks are followed by a convolutional layer to extract features and make a global residual. A sub-pixel convolution module is used to increase the size of the image. A sub-pixel convolution module increases the size by two times, and finally a convolutional layer is used to obtain a high-resolution image.

[0047] The discriminator model is used to determine whether the original image or the reconstructed image belongs to the generated image, and obtain the final imaging of the target; specifically: for the structure of the discriminator model, the discriminator model judges the original high-definition image or the reconstructed high-definition image to determine whether the image is created by the generator. It is essentially a classification model. Therefore, the final output of the discriminator is a 1-dimensional tensor. The discriminator network is a CNN network in which the activation function uses Leaky-ReLU (α=0.2) to prevent some negative output necrosis. The ending part uses a Dense block plus a sigmoid function for binary classification, that is, to determine whether it is a generated image or an original high-definition image. The overall discriminant network is a VGG network without the pooling layer.

[0048] In the present application, when the weather conditions are good during the day, the detection distance is the result of the combined effect of subjective and objective factors. The equivalent stripes of the target are a set of black and white stripe patterns with equal intervals, whose total height is the critical size of the target, and the stripe length is the horizontal target size of the target perpendicular to the critical size. The resolution of the equivalent stripe pattern is the number of resolvable stripes contained in the critical size of the target, that is, the number of pixels occupied by the target imaging on the detector.

[0049] Detection is defined as: when a target is found in the field of view, the image of the target must occupy more than 1.5 pixels (i.e. 2 pixels) in the critical size direction; when the sky is used as the background, it must occupy more than 0.8 pixels (i.e. 1 pixel).

[0050] Recognition is defined as being able to identify the type of target from its outline, such as a tank, truck, or person. In this case, the image of the target must occupy more than 8 pixels in the critical size direction.

[0051] Index requirements: For typical ground vehicle targets, the detection distance is greater than 2km.

[0052] The target detection distance is: ; In the formula, is the pixel size of the imaging device, is the number of pixels occupied by the detected target, is the focal length of the imaging device, For the actual height, is the actual object distance.

[0053] Assuming that the size of a typical ground vehicle target is: 4m×2.4m, the pixel size of the detector used in the system is 1.45um, and the number of pixels occupied by the target is 1.5, if the detection distance is at least 2km, according to the formula, the system focal length is calculated as: ,but: ; in, is the detector pixel size during detection, is the number of pixels occupied by the detected target, is the actual object height during detection, is the actual object distance during detection.

[0054] For moving target capture, the size of a conventional aerial moving platform target is 1.5m×1.5m. The pixel size of the detector used in the system is 1.45um. When the number of pixels occupied by the target is 8, if the recognition distance is at least 200m, the system focal length is calculated according to the formula: ,but: ; in, is the detector pixel size at the time of capture, is the number of pixels occupied by the captured target, is the actual height of the object when captured, is the actual object distance at the time of capture.

[0055] The above-mentioned bionic compound eye imaging device is designed for low-altitude unmanned platform scene perception tasks (such as aerial surveillance, target detection and recognition, etc.), and a large-field-of-view compact bionic compound eye imaging device is designed. The camera is used to divide the field of view into several areas to achieve seamless division of the predetermined field of view, and the camera arrangement rules are confirmed according to the mapping relationship between the field of view and the camera to achieve the desired field of view. It can meet the performance requirements of an effective field of view of 160° and a detection distance of 2km for typical ground vehicle targets. It has the advantages of large field of view, small distortion and high resolution, and has the post-processing functions of multi-channel image display, image stitching, and image super-resolution. It realizes large-field-of-view imaging and multi-view time-series image super-resolution reconstruction, improves the effective field of view and detection distance of target detection and information capture, improves imaging resolution and image quality, and supports low-altitude visual algorithm research. In addition, this application uses the spherical coordinate system of the spherical shell as the design reference and optical reference to make the structure more compact, and the array camera group is modularly designed to realize the synchronous image acquisition function of the multi-camera module, providing technical support for the low-cost, lightweight and intelligent development of the optoelectronic sensor system of the low-altitude unmanned platform.

[0056] In a specific embodiment, Figure 6 The distribution diagram of the imaging device shown in the figure includes a central unit and eight edge units. According to the simulation and angle calculation of the actual lens field of view angle, one central lens and eight edge lenses are arranged at an angle of 55°. The imaging system can meet the 160° field of view requirement.

[0057] The head cover and tail cover are mainly made of aviation aluminum alloy, which is light and strong. After surface coloring and oxidation treatment, it has good corrosion resistance. The lens of the small camera is made of resin material and injection molded; the structural parts connecting the imaging equipment and the image processing equipment are made of ABS material and molded by mold processing, with good processing accuracy and consistency. The volume and total weight of multiple cameras are very small, and the connected structural parts also further compress the size of the printed circuit board, providing a feasible solution for miniaturized multi-camera arrays.

[0058] The whole machine is sealed by adding O-rings or applying silicone rubber at the edge of the front lens, the flange connection position of the structural parts and the aviation socket position.

[0059] The interface between the front-end camera and the image processing device adopts a J30J aviation plug to power the front-end camera, and 7-channel camera image acquisition is output to the image processing device; the camera power supply voltage is 5V DC voltage; the image processing device and the external interface adopt a J30J aviation plug, and an aviation socket is installed on the rear end cover, through which it can be connected to the outside; the fan power supply is 5V DC voltage.

[0060] like Figure 7 As shown in the size diagram, the diameter of the hood is 118mm and the total length is 139mm. After preliminary calculation, the weight of the whole machine is about 0.9kg.

[0061] For the central unit, the design is carried out according to the technical analysis indicators. The system band is visible light, the field of view design input indicator is 80.2°×51°, the focal length is 3.24mm, the detector resolution is 3840×2160, and the pixel size is 1.45um×1.45um.

[0062] There are seven major geometric aberrations in the optical system, namely spherical aberration, coma, field curvature, astigmatism, chromatic aberration of magnification and position chromatic aberration, and there are also a large number of higher-order aberrations. The purpose of optical design is to balance the relationship between these seven major geometric aberrations, and balance the remaining primary aberrations with the higher-order aberrations, so that the imaging quality of the optical system meets the requirements.

[0063] Based on the decomposed indicators and requirements, the imaging optical lens group of the optical system adopts a three-piece structure. After reasonable matching, the imaging requirements can be achieved and the optical imaging quality can be improved. The optical design is carried out according to the decomposed optical indicators. The schematic diagram of the optical system is shown in the figure below. Figure 8 shown.

[0064] The modulation transfer function reflects the relationship between the spatial frequency of the system and the image contrast. It is a key indicator for evaluating system performance, and mainly considers the MTF value at the Nyquist frequency. MTF represents the degree of contrast attenuation of various sinusoidal intensity distribution functions after imaging by the optical system. Therefore, when a certain frequency drops to 0, it means that the light intensity distribution at this frequency has no brightness change, that is, the frequency is cut off. This is the main method of using the optical transfer function to evaluate the image quality of the optical system. The MTF curve of the optical system at the cutoff frequency is as follows: Fig. 9 shown.

[0065] After many rays of light emitted from a point pass through the optical system, due to aberration, they are no longer focused on the same point as the image plane, but form a diffuse pattern scattered in a certain range, which is called a spot diagram. The method of using the density of these points in the spot diagram to measure the imaging quality of the optical system is called the spot diagram method. For a system with large aberrations, the density of the intersection points of these rays in the spot diagram can be used to measure the quality of its imaging. The spot diagram of the optical system is shown in Figure 1. Fig.10 shown.

[0066] For the edge unit, the edge lens is designed according to the technical analysis indicators. Through the angle position and field of view simulation between the center lens and the edge lens, the same lens as the center is used for design. The system band is also visible light, the field of view design input is 80.2°×51°, and the focal length is 3.24mm. After analysis and optimization, the optical system structure diagram of the edge unit is as follows: Fig.11 shown.

[0067] In order to reduce the size of the system structure, the central unit and the edge unit are designed in the form of lens modules. An integrated development board is selected. The detector resolution is 3840×2160, the pixel size is 1.45um×1.45um, and the USB2.0 output port is used. The overall structure diagram of the module is as follows: Fig.12 As shown, the size diagram is as follows Fig.13 and Fig.14 shown.

[0068] The simulation is performed based on the field of view angles of the center lens and the edge lens. The bird’s-eye view result of the camera field of view simulation is as follows: Fig.15 As shown, the side view result of the camera field of view stitching simulation is as follows Fig.16 As shown in the splicing simulation diagram, it can be seen that the overlap area between the center field of view camera and the edge field of view camera is relatively large, and the entire 160° field of view area can be covered.

[0069] like Fig.17The stitched panoramic image is shown below. Taking the stitched panoramic image as an example, a partial area at the position shown by the lower square is taken as the super-resolution area of the image. The image of the selected area is intercepted and input into the super-resolution network. The super-resolution algorithms SRGAN or EDSR are selected to perform super-resolution operations of ×2 or ×4 on the selected area, and the super-resolution image result is as shown Fig.18 shown below.

[0070] After testing, the developed large field of view compact bionic compound eye imaging and image processing system meets the following indicators: (1) Platform used: low-altitude unmanned platform, typical cruising altitude 200 - 300m; (2) Array size: 40mm < R < 60mm, hemispherical surface; (3) Payload weight: < 1kg; (4) Effective field of view: ≥160°; (5) Detection distance: for typical ground vehicle targets, the detection distance is greater than 2km; (6) Image processing:具备任务和场景驱动的宽视场多分辨图像重建能力。

[0071] The content not described in detail in this specification belongs to the prior art well-known to those skilled in the art.

[0072] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0073] The above-described embodiments merely represent several implementation manners of the present application, and their descriptions are relatively specific and detailed, but should not be construed as limiting the scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A bionic compound eye imaging device, characterized in that: include: A hood and an imaging device disposed on the hood; The head cover is a hemispherical structure, the interior of the hemispherical structure is hollow, forming a shell space for arranging an image processing device; The imaging device comprises: a central unit arranged at the top of the hood and a plurality of edge units arranged in an interval array around the central unit to form a bionic compound eye structure; the central unit and the edge units are both connected to the image processing device to achieve target detection imaging.

2. The bionic compound eye imaging device according to claim 1, characterized in that: The number of the edge units is eight, and they form a circular array along the circumference of the head cover.

3. A bionic compound eye imaging device according to claim 1 or 2, characterized in that: The central unit and the edge unit both include: a lens module and a base; The lens module corresponds to the base one by one, and the lens module is arranged on the head cover by corresponding to the base.

4. The bionic compound eye imaging device according to claim 3, characterized in that: The lens module adopts a three-piece structure.

5. The bionic compound eye imaging device according to claim 4, characterized in that: The optical path of the lens module is perpendicular to the section corresponding to the head cover position where the base is located.

6. A bionic compound eye imaging device according to claim 1 or 2, characterized in that: Also includes: Tail cover and heat dissipation equipment; The tail cover is a circular ring structure, the diameter of which is equal to the diameter of the hemispherical structure, so that one end of the tail cover is connected to the edge of the head cover; The heat dissipation device is arranged in the inner space of the tail cover.

7. A bionic compound eye imaging device according to claim 1 or 2, characterized in that: The target detection distance is: In the formula, is the pixel size of the imaging device, is the number of pixels occupied by the detected target, is the focal length of the imaging device, For the actual height, is the actual object distance.

8. A bionic compound eye imaging device according to claim 1 or 2, characterized in that: The image processing device comprises: an image stitching component and an image super-resolution component, which are used for image stitching and image super-resolution reconstruction respectively.

9. The bionic compound eye imaging device according to claim 8, characterized in that: The image stitching component includes: an image preprocessing module, an image registration module and an image fusion module connected in sequence; The image preprocessing module is used to reconstruct the acquired original image by using a histogram specification method and a histogram equalization method to obtain a preprocessed image; The image registration module is used to use a feature-based registration method to perform feature extraction, feature matching and perspective transformation matrix calculation on the pre-processed image to obtain a registered image; The image fusion module is used to fuse the registered images using a weighted average fusion algorithm to obtain a stitched image.

10. The bionic compound eye imaging device according to claim 9, characterized in that: The image super-resolution component includes: a generator model and a discriminator model connected; The generator model is used to reconstruct the spliced ​​image to obtain a reconstructed image; The discriminator model is used to determine whether the original image or the reconstructed image belongs to the generated image, and obtain the final imaging of the target.

Images