A compound eye image extraction and map construction system for port SLAM

Abstract

A compound eye image extraction and map construction system for port SLAM includes: a compound eye visual SLAM subsystem, a DSP digital image processing module and a computer subsystem. The DSP digital image processing module digitally processes the map image constructed by the compound eye visual SLAM subsystem and transmits it to the computer subsystem. The present invention adopts compound eye image extraction on the basis of the visual SLAM subsystem, extracts and collects data through the visual SLAM subsystem, and then transmits the data to the DSP digital image processing module. The collected data is processed by the computer's image processing module, and finally extracted and displayed by the human-computer interface module. The image extraction and feature matching method of the compound eye combined with the visual SLAM subsystem is adopted to solve the data association problem in SLAM. The extracted image has different expressions of different regions. The number of feature points should be much smaller than the number of pixels. Compared with the traditional multi-eye image extraction method, the image extraction efficiency and effect are improved.

Description

Technical Field

[0001] The invention relates to the technical field of image extraction, and in particular to a compound eye image extraction and map construction system for port SLAM. Background Art

[0002] In some areas of the port, people cannot directly enter the space to observe the internal structure due to spatial and geographical limitations or certain risks to the safety of its location. At this time, robots are needed to build and locate the map in real time to achieve image extraction and map construction in unknown areas. When the existing visual SLAM image is extracted, the distinctiveness and efficiency of the extracted image features are low, and data association problems will occur. When the existing multi-eye combination method is used to extract images, its clarity, visual mileage, field of view and other effects are relatively poor, and the response time is also long. Summary of the invention

[0003] The object of the present invention is to provide a compound eye image extraction and map construction system for port SLAM.

[0004] The specific technical solutions of the present invention are as follows:

[0005] A compound eye image extraction and map construction system for port SLAM includes: a compound eye vision SLAM subsystem, a DSP digital image processing module and a computer subsystem. The DSP digital image processing module digitally processes the map image constructed by the compound eye vision SLAM subsystem and transmits it to the computer subsystem.

[0006] Optionally, the compound eye vision SLAM subsystem includes:

[0007] At least one compound eye vision sensor component, used to collect environmental images within the collection range around the compound eye vision sensor component and perform preliminary processing on the images;

[0008] The SLAM module is used to perform secondary processing on the images collected by the compound eye vision sensor component and construct a map.

[0009] Optionally, the compound eye vision sensor assembly includes: a mounting frame, a plurality of image acquisition units, a plurality of sub-DSP processing modules and a plurality of image comparators;

[0010] The mounting frame includes a telescopic arm and a pan-tilt mounted on the telescopic arm, a plurality of mounting grids are arranged in a circular array on the outer peripheral side of the pan-tilt with the center of the pan-tilt as the center, a plurality of image acquisition units are respectively arranged in the mounting grids, and the plurality of image acquisition units are arranged in a ring structure to collect scene information such as environmental images within a 360° range around the pan-tilt as the center;

[0011] Each of the image acquisition units is electrically connected to a sub-DSP processing module, which is installed on the pan-tilt platform. The sub-DSP processing module is used to perform preliminary processing on the acquired scene information, and two adjacent image acquisition units are electrically connected to an image comparator respectively.

[0012] Optionally, the compound eye vision sensor assembly further includes: an environment sensor module and a robot body sensor module;

[0013] The environmental sensor module includes: a GPS positioning unit and a reconstruction model unit, the GPS positioning unit and the reconstruction model unit are arranged on the mounting frame, the GPS positioning unit is configured to obtain the geographical coordinates of the compound eye vision sensor assembly and the collected image; the reconstruction model unit is used to obtain the position of the compound eye vision sensor assembly according to the physical space of the target object and the accuracy of the reconstruction model planned by the target object collection network;

[0014] The robot body sensor includes an inertial measurement unit, which is used to detect the three-axis attitude angle and acceleration of the compound eye vision sensor assembly.

[0015] Optionally, the compound eye image extraction and map building system for port SLAM also includes a positioning subsystem, which includes: multiple base stations and at least one tag, each of the base stations has a unique number, and the multiple base stations are distributed in the area to be detected. The tag is set on the mounting frame of the compound eye vision sensor assembly, and each of the compound eye vision sensor assembly corresponds to a unique tag. The numbers of different base stations are recorded in the tag, and the positioning subsystem transmits a transmission pulse of a preset frequency according to the tag, and the base station receives the pulse emitted by the tag and calculates the distance from the tag to the base station.

[0016] Optionally, the SLAM module includes: a sensor data unit, a front-end data odometer unit, a back-end nonlinear optimization unit, a loop detection unit and a mapping unit; the sensor data unit, the front-end data odometer unit, the back-end nonlinear optimization unit, and the mapping unit are electrically connected in sequence, one end of the loop detection unit is electrically connected to the sensor data unit, and the other end of the loop detection unit is electrically connected to the back-end nonlinear optimization unit.

[0017] Optionally, the computer subsystem includes: an image processing module, which is used to process the digital image information transmitted by the DSP digital image processing module into analog image information.

[0018] Optionally, the image processing module includes: an image preprocessing unit, an image segmentation unit, an image morphology processing unit and an image operation unit.

[0019] Optionally, the computer subsystem further includes: a human-machine interface module, and the human-machine interface module is used to display the processed image.

[0020] Optionally, the DSP digital image processing module includes: a control processor, a DSPs unit, a data transmission unit, a memory and an input / output interface.

[0021] The beneficial effect of the present invention lies in that the present invention adopts compound eye image extraction on the basis of the visual SLAM subsystem, extracts and collects data through the visual SLAM subsystem, and then transmits the data to the DSP digital image processing module. The collected data is processed by the image processing module of the computer, and finally extracted and displayed by the human-computer interface module. The image extraction and feature matching method of the compound eye combined with the visual SLAM subsystem is adopted to solve the data association problem in SLAM. The extracted image has different areas with different expressions. The number of feature points should be much smaller than the number of pixels. Compared with the traditional multi-eye image extraction method, the image extraction efficiency and effect are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a schematic diagram of the system block diagram of the present invention;

[0023] Figure 2 This is a schematic diagram of the mounting frame structure of the present invention;

[0024] Figure 3 This is a schematic diagram of the top view of the mounting frame of the present invention;

[0025] Figure 4 This is a schematic diagram of the system block diagram of the compound eye vision sensor assembly of the present invention;

[0026] Figure 5 It is a schematic diagram of the positioning subsystem of the present invention;

[0027] Figure 6 This is a schematic diagram of the SLAM module block diagram of the present invention;

[0028] Figure 7 This is a schematic diagram of a block diagram of an image processing module of the present invention;

[0029] Figure 8 This is a schematic diagram of a DSP digital image processing module block diagram of the present invention;

[0030] Fig. 9 This is a schematic diagram of the distribution of image acquisition units of the present invention;

[0031] Fig.10 This is a schematic diagram of overlapping visual field areas of the image acquisition unit of the present invention;

[0032] Fig.11A regional growth flow chart of the grayscale characteristics of the system is constructed for the present invention. DETAILED DESCRIPTION

[0033] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0034] like Figure 1 As shown, the present invention provides a compound eye image extraction and map construction system for port SLAM, including: a compound eye vision SLAM subsystem 1, a DSP digital image processing module 2 and a computer subsystem 3. The DSP digital image processing module 2 digitally processes the map image constructed by the compound eye vision SLAM subsystem 1 and transmits it to the computer subsystem 3.

[0035] The compound eye vision SLAM subsystem 1 includes:

[0036] At least one compound eye vision sensor component 11, used to collect environmental images within the collection range around the compound eye vision sensor component 11 and perform preliminary processing on the images;

[0037] The SLAM module 12 is used to perform secondary processing on the images collected by the compound eye vision sensor assembly 11 and construct a map.

[0038] The computer subsystem 3 includes: an image processing module 31, which is used to process the digital image information transmitted by the DSP digital image processing module 2 into analog image information.

[0039] like Figure 2-4 As shown, the compound eye visual sensor assembly 11 includes: a mounting frame 111, multiple image acquisition units 112, multiple sub-DSP processing modules 113, and multiple image comparators (not shown in the figure). The mounting frame 111 includes a telescopic arm 1111 and a pan-tilt 1112 arranged on the telescopic arm 1111. The outer peripheral side of the pan-tilt 1112 is provided with multiple mounting grids 1113 in a circular array with the center of the pan-tilt 1112 as the center. The multiple image acquisition units 112 are respectively arranged in the mounting grids 1113. The multiple image acquisition units 112 are arranged in a ring structure to collect scene information such as environmental images within a 360° range around the pan-tilt 1112 as the center, which can form a large field of view of 360° and realize the compound eye parallel processing mode.

[0040] like Figure 4As shown, each of the image acquisition units 112 is electrically connected to a sub-DSP processing module 113, and the sub-DSP processing module 113 is installed on the pan-tilt head 1112. The sub-DSP processing module 113 is used to perform preliminary processing on the collected scene information to form signal acquisition and preprocessing. Two adjacent image acquisition units 112 are electrically connected to an image comparator respectively.

[0041] The image acquisition unit 112 is a CCD camera, which is used to collect scene information such as environmental images. The length of the telescopic arm 1111 is adjustable, and the length of the telescopic arm 1111 is adjustable from 300mm to 600mm. The height of the plane of the pan-tilt head 1112 can be adjusted to facilitate observation of the surrounding environment.

[0042] In order to ensure that the compound eye vision sensor component 11 can collect, process and compare information in parallel, each CCD camera of the present invention is connected to a sub-DSP processing module 113, which is responsible for the information processing of its own scene, and there is an image comparator between each two adjacent CCD cameras to realize the comparison and processing of the image content. In this way, no matter whether it is information collection, processing or comparison, parallel processing can be achieved.

[0043] When the compound eye vision sensor component 11 extracts images, a circular array of image acquisition units 112 is used to capture the picture, forming a compound eye acquisition effect. The compound eye vision sensor component 11 digitizes the technical picture and then binarizes and filters the images with the same grayscale level to combine and analyze them, thereby achieving image extraction within a 360-degree range. Compared with the multi-eye multi-group combination method, it saves equipment installation.

[0044] The compound eye vision sensor assembly 11 further includes: an environment sensor module and a robot body sensor module (not shown).

[0045] The environmental sensor module includes: a GPS positioning unit and a reconstruction model unit, the GPS positioning unit and the reconstruction model unit are arranged on the mounting frame 111, and the GPS positioning unit is configured to obtain the geographical coordinates of the compound eye visual sensor assembly 11 and the collected image. The reconstruction model unit is used to obtain the position of the compound eye visual sensor assembly 11 according to the target object acquisition network planning target object's physical space and the accuracy of the reconstruction model. The boundary of the target object is determined by the reconstruction model unit, and the target object is quickly reconstructed, which can ensure that the reconstruction is a three-dimensional scene of the same time section, and the real-time performance of the three-dimensional scene is guaranteed.

[0046] The robot body sensor includes an inertial measurement unit, which is used to detect the three-axis attitude angle and acceleration of the compound eye vision sensor assembly 11.

[0047] like Figure 5 As shown, the compound eye image extraction and map building system for port SLAM also includes a positioning subsystem 4, which includes: multiple base stations 41 and at least one tag 42, each of which has a unique number, and multiple base stations 41 are distributed in the area to be detected. The tag 42 is set on the mounting frame 111 of the compound eye visual sensor assembly 11, and each compound eye visual sensor assembly 11 corresponds to a unique tag 42. The numbers of different base stations 41 are recorded in the tag 42. The positioning subsystem 4 transmits a transmission pulse of a preset frequency according to the tag 42, and the base station 41 receives the pulse emitted by the tag 42 and calculates the distance from the tag 42 to the base station 41. The base station 41 can radiate the area where the target object to be photographed is located. The distance from the tag 42 to the base station 41 calculated by receiving the tag 42 pulse can be used for the precise positioning of the compound eye visual sensor assembly 11 in the target object acquisition network.

[0048] By acquiring accurate compound eye vision sensor components 11 and image geographic coordinates through multiple base stations 41, the target object is processed in a networked manner and the compound eye vision sensor components 11 are dispatched for shooting, which can reduce the amount of data processing.

[0049] like Figure 1 , 6As shown, the SLAM module 12 includes: a sensor data unit 121, a front-end data odometer unit 122, a back-end nonlinear optimization unit 123, a loop detection unit 124 and a mapping unit 125. The sensor data unit 121, the front-end data odometer unit 122, the back-end nonlinear optimization unit 123, and the mapping unit 125 are electrically connected in sequence, one end of the loop detection unit 124 is electrically connected to the sensor data unit 121, and the other end of the loop detection unit 124 is electrically connected to the back-end nonlinear optimization unit 123. The SLAM module 12 can adopt an existing conventional module structure. In the visual SLAM problem, the estimation of the position is often a recursive process, that is, the current frame position is solved by the previous frame position, so the error therein is passed down frame by frame, that is, the cumulative error. An effective way to eliminate the error is to perform loop detection. Loop detection determines whether the robot has returned to the position previously passed. If a loop is detected, it will pass the information to the back end for optimization processing. The loop is a more compact and accurate constraint than the backend, and this constraint can form a topologically consistent trajectory map. If the closed loop can be detected and optimized, the result can be made more accurate. The entire SLAM can be roughly divided into the front end and the back end. The front end is equivalent to VO (visual odometer), which studies the transformation relationship between frames. First, extract the feature points of each frame image, use the adjacent frame images, match the feature points, and then use RANSAC to remove the large noise, and then match to obtain a pose information (position and attitude). At the same time, the attitude information provided by IMU (Inertial measurement unit) can be used for filtering and fusion. The back end mainly optimizes the results of the front end, and uses filtering theory (EKF, UKF, PF), or optimization theory TORO, G2O to optimize the tree or graph. Finally, the optimal pose estimation is obtained. Because it is based on the filtering theory, the filter stability grows too fast, which puts a lot of pressure on the EKF (Extended Kalman Filter) that needs to be frequently inverted. Graph-based SLAM usually uses keyframes as the basis to establish relative transformation relationships between multiple nodes, such as affine transformation matrices, and continuously maintains key nodes to ensure the capacity of the graph, thereby reducing the amount of calculation while ensuring accuracy.

[0050] like Figure 1 , 7 As shown, the image processing module 31 includes: an image preprocessing unit 311, an image segmentation unit 312, an image morphology processing unit 313 and an image operation unit 314.

[0051] The image preprocessing unit 311 processes the input image before feature extraction, segmentation and matching. The main purpose is to eliminate irrelevant information in the image, restore useful real information, enhance the detectability of relevant information and simplify data to the maximum extent, thereby improving the reliability of feature extraction, image segmentation, matching and recognition.

[0052] The image segmentation unit 312 divides the image into several specific regions with unique properties and proposes the technology and process of the target of interest. It is a key step from image processing to image analysis. The existing image segmentation methods are mainly divided into the following categories: threshold-based segmentation methods, region-based segmentation methods, edge-based segmentation methods, and segmentation methods based on specific theories. From a mathematical point of view, image segmentation is the process of dividing a digital image into non-intersecting regions. The image segmentation process is also a labeling process, that is, pixels belonging to the same region are assigned the same number.

[0053] The image morphology processing unit 313 refines and differentiates data features, including dilation and erosion, refinement and coarsening, and the like.

[0054] The image operation unit 314 performs arithmetic operations, logical operations and mathematical morphology operations on the image.

[0055] like Figure 1 As shown, the computer subsystem 3 further includes: a human-machine interface module 32, and the human-machine interface module 32 is used to display the processed image.

[0056] like Figure 1 , 8 As shown, the DSP digital image processing module 2 includes: a control processor 21, a DSPs unit 22, a data transmission unit 23, a memory 24 and an input / output interface 25. The control processor 21 is used to complete the control functions of the system including host command interpretation, data transmission control and data input / output. The control processor 21 can enable the DSPs unit 22 to focus on the implementation of high-speed real-time DSP algorithms. The DSPs unit 22 completes the real-time signal processing algorithm. The data transmission unit 23 is used to realize the interconnection between various modules to transmit data. The memory 24 supports data storage, and its main parameter indicators include the capacity of the memory 24, the word length of the memory 24, the access speed, the support for special addressing modes and the storage management control capability. The input / output interface 25 is used to input the results to be processed or processed. The main parameters of the input / output interface 25 include interface bandwidth, buffer storage capacity, data word length, interface protocol and the number of inputs and outputs supported by the interface.

[0057] The DSP digital image processing module 2 can preferably adopt a high-performance fixed-point DSP digital image processing module 2 of model TMS320C6416, whose computing power can reach 4Gips, and can perform real-time video, audio and picture acquisition, compression and playback. The operating frequency can be 600MHZ, the external bus is EMIFA and EMIF, and the bus clock is 100MHZ. The 2.2 version of the PCI interface is used as the host interface.

[0058] The specific principles of the present invention are as follows:

[0059] like Figure 9-10 As shown, during the acquisition, the position information of the feature point is obtained in the following way: first, the radius of the circle where the CCD camera is located is L; the diameter of the CCD camera is D; half of the horizontal viewing angle of the CCD camera is β; the angle between the center lines of the visual axes of adjacent CCD cameras is α,

[0060] The overlapping area of ​​the field of view of two adjacent CCD cameras is a quadrilateral area M, which is the intersection area of ​​the field of view of two adjacent CCD detectors. The area observed by only these two CCD cameras has only an angle difference. The areas observed by different numbers of CCD cameras have a radius difference in the visual axis direction, which is represented by R. i Indicates the radius of the area observed by the i-th CCD camera. The areas observed by different combinations of CCD cameras have different angles on the visual axis tangent. θ(α) is used to represent the angle of the area observed by different combinations of CCD cameras. The parameters of the compound eye vision SLAM subsystem 1 are L (the radius of the circle where the CCD camera is located is L), α (the angle between the center lines of the visual axis of adjacent CCD cameras), β (half of the horizontal viewing angle of the CCD camera), N (the number of cameras), p, q (represented by a matrix with position information, whose values ​​are fixed), and the distance R i and circumferential resolution θ(α) to represent the spatial position of the moving target at a certain moment. Then θ(α) and R i It can be obtained by the following formula:

[0061] y = tan β(xL);

[0062] y=tan β((i-1)α-β)[xL cos((i-1)α)]+L sin((i-1)α);

[0063] By calculation, we know that the coordinates of the point are:

[0064]

[0065]

[0066] So the distance to the target point is:

[0067] where i = 1, 2, … [2β / α];

[0068]

[0069] In this way, the position information of the feature points can be represented by the compound eye vision sensor component 11 and the DSP digital image processing module 2.

[0070] After the compound eye vision SLAM subsystem 1 extracts the image, the DSP digital image processing module 2 converts the G, R, B ternary color space of the image into grayscale. The grayscale composition of the image is: C = xR + yG + zB, where x + y + z = 1, x, y, z ∈ [0, 1]. In the image grayscale conversion, let x, y, z be 0.299, 0.587, and 0.114 respectively. Then, the color image is converted into a grayscale image through the conversion formula system. The conversion method is:

[0071]

[0072] A color image is converted into a grayscale image by the operation of this formula. Next, the DSPs unit 22 in the DSP digital image processing module 2 is used to digitize the image. The image digitization will be divided into two steps. First, the system will perform sampling, expand the image function under a set of orthogonal bases, perform Fourier transform, and use the expanded coefficients to represent the digitized image. If each row (horizontal) pixel is M and the column pixel is N, the image size is M×N pixels, so f(x, y) constitutes an M×N real number matrix:

[0073]

[0074] Each element is a discrete sampling value of the image f(x, y) as a pixel. The result of sampling is the grayscale value of each pixel. Secondly, the system quantizes the continuous value (brightness) of the sampled image function based on sampling, converting it into its digital equivalent. For an image with M×N pixels, the number of grayscale binary bits of each pixel after quantization is Q. Generally, Q is an integer power of 2, that is, Q=2 K , then the number of binary bits b required to store a digital image is:

[0075] b = M × N × Q;

[0076] The number of bytes B is

[0077] B = M × N × Q / 8;

[0078] After the image is digitized, the gray value of each pixel is an integer, and the processing result is a two-dimensional integer matrix, that is, a digital image. The DSP digital image processing module 2 temporarily stores the digital image in a register to be called when the image is spliced.

[0079] After the image is digitized, the DSPs unit 22 will perform binarization, that is, convert a graphic with multiple gray levels into an image with only two gray levels, so as to facilitate data compression, target prominence and feature target recognition. Before constructing the surrounding environment, a threshold T must be specified through the environment, and T is used to divide the image data into two parts: a pixel group greater than T and a pixel group less than T. If the gray value of a pixel in the image is less than the threshold, the gray value of the pixel is set to 0, otherwise the gray value is set to 255.

[0080] In the actual bitmap, the RGB value corresponding to 0 is 0, and the RGB value corresponding to 1 is 255. The binarization process is processed according to the following formula:

[0081]

[0082] After threshold processing, the image becomes a black and white binary image.

[0083] The original image will be disturbed by various noises during the acquisition and transmission process, which will reduce the image quality. The nonlinear median filter method is used to eliminate noise, which can maintain the details in the image and prevent edge blurring. It first determines an odd pixel window W, sorts the pixels in the window from small to large according to the grayscale value, and replaces the original grayscale value with the grayscale value at the middle position. Assuming that the grayscale value of the enhanced image at (x, y) is f(x, y), and the grayscale value of the enhanced image at the corresponding position (x, y) is g(x, y), then:

[0084] g(x,y)=median{f(xk,yl),k,l∈W};

[0085] Median filtering allows pixels with grayscale values ​​that are significantly different from those of surrounding pixels to take values ​​that are close to the grayscale values ​​of surrounding pixels, thus eliminating noise.

[0086] like Fig.11 As shown, after the DSP digital image processing module 2 binarizes the image, the digital image is transmitted to the image processing module 31 via the memory 24 through the output interface, and the image processing module 31 performs region growing on the digital image to form a combined image, and the specific steps are as follows:

[0087] Step 1. Set the mode, use the image transmitted by the CCD camera of the first image (which can be set as 1) as the reference to sequentially scan the images, find a boundary pixel point, and set the pixel as (X0, Y0);

[0088] Step 2. With (X0, Y0) as the center, check the neighboring pixel (X, Y) of (X0, Y0) to determine whether (X, Y) satisfies the growth rule. If so, merge (X, Y) with (X0, Y0) and push (X, Y) into the stack. If not, discard the neighboring pixel (X, Y).

[0089] Step 3. Take the pixel boundary from the stack as (X0, Y0) and go back to step 2.

[0090] Step 4. When the stack is empty, go back to step 1.

[0091] Step 5. Repeat step 1 or step 4 until every pixel in the image has an attribute, and the growth is completed.

[0092] When the first image cannot grow normally and the data is lost, the image processing module 31 will search for feature points through multiple images, that is, determine the correspondence between the current road sign and the road sign seen before, and then perform feature matching. The simplest feature matching method is brute force matching, that is, measure the distance of the descriptor for each feature point and all feature points, then sort them, and take the nearest one as the matching point. The distance of the descriptor expresses the similarity between the two features. Then grow again with the matched feature points as the boundary, which solves the data association problem.

[0093] The compound eye image extraction effect and multi-eye image extraction effect adopted by the present invention are compared with the traditional single eye image extraction, as shown in Table 1 below:

[0094]

[0095] It can be seen from the above table that the clarity, visual mileage, field of view and reaction time of compound eye image extraction are more ideal than those of multi-eye image extraction.

[0096] The beneficial effect of the present invention lies in that the present invention adopts compound eye image extraction on the basis of the visual SLAM subsystem, extracts and collects data through the visual SLAM subsystem, and then transmits the data to the DSP digital image processing module. The collected data is processed by the image processing module of the computer, and finally extracted and displayed by the human-computer interface module. The compound eye structure is combined with the image extraction and feature matching method of the visual SLAM subsystem to solve the data association problem in SLAM. The extracted image has different areas with different expressions. The number of feature points should be much smaller than the number of pixels. Compared with the traditional multi-eye image extraction method, the image extraction efficiency and effect are improved.

[0097] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A compound eye image extraction and map construction system for port SLAM, characterized in that: include: A compound eye vision SLAM subsystem (1), a DSP digital image processing module (2) and a computer subsystem (3), wherein the DSP digital image processing module (2) digitizes the map image constructed by the compound eye vision SLAM subsystem (1) and transmits the digitized map image to the computer subsystem (3); The compound eye vision SLAM subsystem (1) comprises: At least one compound eye visual sensor component (11) is used to collect environmental images within a collection range around the compound eye visual sensor component (11) and perform preliminary processing on the images; A SLAM module (12) is used to perform secondary processing on the images collected by the compound eye vision sensor assembly (11) and construct a map; The compound eye visual sensor assembly (11) comprises: a mounting frame (111), a plurality of image acquisition units (112), a plurality of sub-DSP processing modules (113), and a plurality of image comparators; The mounting frame (111) comprises a telescopic support arm (1111) and a pan / tilt platform (1112) arranged on the telescopic support arm (1111); a plurality of mounting grids (1113) are arranged in a circular array on the outer peripheral side of the pan / tilt platform (1112) with the center of the pan / tilt platform (1112) as the center of the circle; a plurality of image acquisition units (112) are respectively arranged in the mounting grids (1113); the plurality of image acquisition units (112) are arranged in a circular structure to acquire scene information such as environmental images within a 360° range around the pan / tilt platform (1112); Each of the image acquisition units (112) is electrically connected to a sub-DSP processing module (113), the sub-DSP processing module (113) is installed on the pan / tilt platform (1112), the sub-DSP processing module (113) is used to perform preliminary processing on the acquired scene information, and two adjacent image acquisition units (112) are electrically connected to an image comparator respectively; The compound eye visual sensor assembly (11) further comprises: an environment sensor module and a robot body sensor module; The environmental sensor module comprises: a GPS positioning unit and a reconstruction model unit, the GPS positioning unit and the reconstruction model unit being arranged on the mounting frame (111), the GPS positioning unit being configured to obtain the geographical coordinates of the compound eye visual sensor assembly (11) and the collected image; the reconstruction model unit being used to obtain the position of the compound eye visual sensor assembly (11) according to the physical space of the target object and the accuracy of the reconstruction model planned by the target object collection network; The robot body sensor comprises an inertial measurement unit, which is used to detect the three-axis attitude angle and acceleration of the compound eye visual sensor component (11); The SLAM module (12) comprises: a sensor data unit (121), a front-end data odometer unit (122), a back-end nonlinear optimization unit (123), a loop detection unit (124) and a mapping unit (125); the sensor data unit (121), the front-end data odometer unit (122), the back-end nonlinear optimization unit (123) and the mapping unit (125) are electrically connected in sequence, one end of the loop detection unit (124) is electrically connected to the sensor data unit (121), and the other end of the loop detection unit (124) is electrically connected to the back-end nonlinear optimization unit (123).

2. The compound eye image extraction and map construction system for port SLAM according to claim 1, characterized in that: The compound eye image extraction and map building system for port SLAM also includes a positioning subsystem (4), the positioning subsystem (4) including: a plurality of base stations (41) and at least one tag (42), each of the base stations (41) having a unique number, the plurality of base stations (41) being distributed in a to-be-detected area, the tag (42) being arranged on a mounting frame (111) of the compound eye visual sensor assembly (11), each of the compound eye visual sensor assembly (11) corresponding to a unique tag (42), the tags (42) recording the numbers of different base stations (41), the positioning subsystem (4) transmitting a transmission pulse of a preset frequency according to the tag (42), the base station (41) receiving the pulse transmitted by the tag (42) and calculating the distance from the tag (42) to the base station (41).

3. The compound eye image extraction and map construction system for port SLAM according to claim 1, characterized in that: The computer subsystem (3) comprises: an image processing module (31) for processing the digital image information transmitted by the DSP digital image processing module (2) into analog image information.

4. The compound eye image extraction and map construction system for port SLAM according to claim 3, characterized in that: The image processing module (31) comprises: an image preprocessing unit (311), an image segmentation unit (312), an image morphology processing unit (313) and an image operation unit (314).

5. The compound eye image extraction and map construction system for port SLAM according to claim 1, characterized in that: The computer subsystem (3) further comprises: a human-machine interface module (32), wherein the human-machine interface module (32) is used to display the processed image.

6. The compound eye image extraction and map construction system for port SLAM according to claim 1, characterized in that: The DSP digital image processing module (2) comprises: a control processor (21), a DSPs unit (22), a data transmission unit (23), a memory (24) and an input / output interface (25).

Images