A sensing method and system for amphibious unmanned boats

Abstract

The present invention provides a sensing method and system for an amphibious unmanned boat, and relates to the technical field of unmanned boat control. The sensing method for an amphibious unmanned boat includes: determining abnormal pixels according to the comparison result of the grayscale value of each pixel in the image and the corresponding adjacent pixel; determining the deviation value corresponding to each abnormal pixel according to each abnormal pixel; determining the abnormal pixel whose deviation value is less than the first preset threshold as an edge pixel, and dividing the image into sub-region images according to all edge pixels; inputting the sub-region images into a trained image classification model to output the region type; determining the target sub-regions passed by the preset driving route when crossing the target region and the order of passing the target sub-regions, and generating the control strategy of the unmanned boat according to the region type corresponding to the target sub-region and the order of passing the target sub-regions. The present invention can effectively improve the adaptability and safety of the unmanned boat control.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned boats, and in particular to a sensing method and system for amphibious unmanned boats. Background Art

[0002] With the advancement of science and technology, the development of unmanned boat technology is facing unprecedented opportunities. Unmanned boats can be used flexibly in a variety of environments, adapt to different driving conditions, and have a wide range of uses. They play an important role in ocean monitoring, polar exploration, and environmental protection.

[0003] However, the control complexity of unmanned boats is relatively high, and different environmental conditions have different control requirements. For example, there are significant differences in the control requirements when driving in different areas such as water, ice and snow. At present, the existing unmanned boat control methods are relatively simple and difficult to adapt to diverse environmental conditions, resulting in poor adaptability and possible loss of control under certain conditions, and insufficient safety. Therefore, for polar environments, by enhancing perception capabilities and combining perception technology to ultimately determine a more effective control strategy, it will help improve the adaptability and safety of unmanned boats in complex environments. Summary of the invention

[0004] The problem solved by the present invention is how to improve the adaptability and safety of unmanned boat operations.

[0005] In order to solve the above problems, the present invention provides a perception method and system for amphibious unmanned boats.

[0006] In a first aspect, the present invention provides a sensing method for an amphibious unmanned boat, comprising:

[0007] Acquire an image of a target area and a preset driving route of the unmanned boat in the target area;

[0008] Determine abnormal pixels in the image based on a comparison result of the grayscale values ​​of each pixel in the image and corresponding adjacent pixels;

[0009] Selecting one of the abnormal pixel points as a reference point, and obtaining a deviation value corresponding to the selected abnormal pixel point according to a distance relationship between the reference point and each of the abnormal pixel points;

[0010] Determine the abnormal pixel points whose deviation values ​​are less than a first preset threshold as edge pixel points, and divide the image into regions according to all the edge pixel points to obtain at least one sub-region image, wherein each sub-region image corresponds to a sub-region of the target region;

[0011] Inputting the sub-region image into a trained image classification model, and outputting the region type corresponding to the sub-region image;

[0012] Determine the target sub-areas that the preset driving route passes through when crossing the target area and the order of passing through the target sub-areas, and generate a control strategy for the unmanned boat according to the area type corresponding to the target sub-area and the order of passing through the target sub-areas.

[0013] Optionally, determining abnormal pixels in the image according to a comparison result of the grayscale values ​​of each pixel in the image and corresponding adjacent pixels includes:

[0014] For a pixel point, determine the grayscale value difference between the pixel point and each of the adjacent pixel points to obtain the adjacent grayscale difference corresponding to each of the adjacent pixel points, wherein the adjacent pixel points are pixel points adjacent to the pixel point;

[0015] Determine the maximum adjacent grayscale difference, the minimum adjacent grayscale difference and the average adjacent grayscale difference according to all the adjacent grayscale differences;

[0016] Obtaining an abnormality score corresponding to the pixel point according to the maximum adjacent grayscale difference, the minimum adjacent grayscale difference and the average adjacent grayscale difference;

[0017] The pixel point whose abnormality score is greater than a second preset threshold is determined as the abnormal pixel point.

[0018] Optionally, the anomaly score satisfies:

[0019] ;

[0020] Where S is the abnormality score, X max is the maximum adjacent grayscale difference, X min is the minimum adjacent grayscale difference, is the average adjacent grayscale difference.

[0021] Optionally, dividing the image into regions according to all the edge pixels to obtain at least one sub-region image includes:

[0022] According to all the edge pixels, a boundary curve is obtained by curve fitting;

[0023] The image is divided according to the boundary curve to obtain the sub-region image.

[0024] Optionally, obtaining a boundary curve by curve fitting according to all the edge pixels includes:

[0025] Generating at least one edge pixel point set according to the distribution of all the edge pixel points;

[0026] According to the edge pixel point set, a plurality of fitting curves are generated by a preset fitting model;

[0027] The fitting curve with the smallest deviation from the edge pixel point set is determined as the boundary curve.

[0028] Optionally, inputting the sub-region image into a trained image classification model and outputting the region type corresponding to the sub-region image includes:

[0029] Determine the number of pixels corresponding to each grayscale value according to the grayscale value of each pixel in the sub-region image;

[0030] Determine the grayscale value ratio corresponding to each grayscale value according to the ratio of the number of pixels corresponding to each grayscale value to the number of all pixels;

[0031] Adding the product of each grayscale value and the corresponding grayscale value ratio to obtain a complementary grayscale value;

[0032] Completing the sub-region image according to the completed grayscale value to generate a sub-region image of a preset size;

[0033] The sub-region image of the preset size is input into the image classification model to obtain the corresponding region type.

[0034] Optionally, the determining the target sub-areas that the preset driving route passes through when crossing the target area and the order of passing through the target sub-areas includes:

[0035] Determine the route pixel points corresponding to the preset driving route in the image;

[0036] Determine a driving sub-region image from among all the sub-region images according to a comparison result between the route pixel points and each of the sub-region images, and determine a sub-region corresponding to the driving sub-region image as the target sub-region;

[0037] The image sequence of the corresponding driving sub-area images is determined according to the sequence of the route pixels in the preset driving route, wherein the image sequence is the sequence in which the preset driving route passes through the target sub-area.

[0038] Optionally, determining the driving sub-region image in the sub-region image according to the comparison result between the route pixel point and each of the sub-region images includes:

[0039] When one of the sub-region images coincides with a pixel point of the route, the sub-region image is determined as the driving sub-region image;

[0040] When a sub-region image does not overlap with a pixel point of the route, the sub-region image is determined as a non-driving sub-region image.

[0041] Optionally, generating the control strategy of the unmanned boat according to the area type corresponding to the target sub-area and the order of passing through the target sub-area includes:

[0042] Determining a driving strategy corresponding to the target sub-area according to the area type corresponding to the target sub-area and a preset matching rule, wherein the matching rule includes a preset one-to-one correspondence between the area type and the driving strategy;

[0043] The determined driving strategies are sorted according to the order in which the preset driving routes pass through the target sub-areas to generate the control strategy of the unmanned boat.

[0044] In a second aspect, the present invention provides a sensing system for an amphibious unmanned boat, comprising:

[0045] An acquisition module, used to acquire an image of a target area and a preset driving route of the unmanned boat in the target area;

[0046] A screening module, used for determining abnormal pixels in the image based on a comparison result of the grayscale values ​​of each pixel in the image and the corresponding adjacent pixels;

[0047] A processing module, configured to select one of the abnormal pixel points as a reference point, and obtain a deviation value corresponding to the selected abnormal pixel point according to a distance relationship between the reference point and each of the abnormal pixel points;

[0048] A comparison module, configured to determine the abnormal pixel points whose deviation value is less than a first preset threshold as edge pixel points, and divide the image into regions according to all the edge pixel points to obtain at least one sub-region image, wherein each of the sub-region images corresponds to a sub-region of the target region;

[0049] A classification module, used for inputting the sub-region image into a trained image classification model and outputting the region type corresponding to the sub-region image;

[0050] A generation module is used to determine the target sub-areas passed by the preset driving route when crossing the target area and the order of passing through the target sub-areas, and generate a control strategy for the unmanned boat according to the area type corresponding to the target sub-area and the order of passing through the target sub-areas.

[0051] In a third aspect, the present invention provides an electronic device, including a memory and a processor;

[0052] The memory is used to store computer programs;

[0053] The processor is used to implement the perception method for amphibious unmanned boats as described in the first aspect when executing the computer program.

[0054] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the perception method for an amphibious unmanned boat as described in the first aspect is implemented.

[0055] The beneficial effects of the perception method and system for amphibious unmanned boats of the present invention are as follows: the pixels in the image of the target area of ​​the unmanned boat are screened, and the abnormal pixels that may be located at the edge of the sub-area image in the image are obtained by comparing the grayscale values ​​of each pixel, and then an abnormal pixel is selected, and the deviation value of the selected abnormal pixel is determined by the distance relationship between the abnormal pixel and all the abnormal pixels, so that the deviation value corresponding to each selected abnormal pixel can be obtained respectively, and the abnormal pixel is further screened according to the comparison result of the deviation value of each abnormal pixel and the first preset threshold, and finally the edge pixel corresponding to the sub-area image in the image is obtained, and the edge position of the corresponding sub-area in the image can be more accurately judged by all the edge pixels, so that the image can be accurately divided into regions according to all the edge pixels to obtain the corresponding sub-area images, each sub-area image corresponds to a sub-area in the target area, and the sub-area images are classified by the trained image classification model to obtain the region type corresponding to the sub-area image, and the driving environment of the sub-area corresponding to the corresponding sub-area image can be obtained according to the region type, so that the appropriate driving strategy can be adjusted in time according to the region type, and the control stability and safety of the unmanned boat in different driving environments are improved. Through the preset driving route, the sub-areas that the unmanned boat passes through when crossing the target area and the order of passing the sub-areas are determined, so that the control strategy of the unmanned boat corresponding to the preset driving route can be generated according to the area type and order corresponding to the sub-areas passed. Through the driving strategy, the unmanned boat is controlled and adjusted accordingly according to different area types, avoiding the inability to adapt to different area types due to a single control method, improving the adjustment ability of the unmanned boat to cope with environmental changes during driving, improving the adaptability of the unmanned boat control, and being able to make timely adjustments and controls to the unmanned boat driving in different sub-areas to meet the control requirements of different types of sub-areas for the unmanned boat, effectively improving the safety of the unmanned boat control. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 A schematic diagram of a flow chart of a sensing method for an amphibious unmanned boat according to an embodiment of the present invention;

[0057] Figure 2 A schematic diagram of image area division according to an embodiment of the present invention;

[0058] Figure 3 This is a schematic diagram of the structure of a sensing system for an amphibious unmanned boat according to an embodiment of the present invention;

[0059] Figure 4 The figure is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0060] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. Although certain embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be interpreted as being limited to the embodiments described herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present invention. It should be understood that the drawings and embodiments of the present invention are only for exemplary purposes and are not intended to limit the scope of protection of the present invention.

[0061] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.

[0062] The term "including" and its variations used in this document are open inclusions, that is, "including but not limited to"; the term "based on" means "based at least in part on"; the term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one other embodiment"; the term "some embodiments" means "at least some embodiments"; the term "optionally" means "optional embodiments". The relevant definitions of other terms will be given in the following description. It should be noted that the concepts of "first", "second", etc. mentioned in the present invention are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0063] It should be noted that the modifications of "one" and "plurality" mentioned in the present invention are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0064] The names of the messages or information exchanged between multiple devices in the embodiments of the present invention are only used for illustrative purposes, and are not used to limit the scope of these messages or information.

[0065] In view of the problems existing in the above-mentioned related technologies, this embodiment provides a perception method and system for amphibious unmanned boats.

[0066] like Figure 1 As shown, an embodiment of the present invention provides a perception method for an amphibious unmanned boat, comprising:

[0067] S100, acquiring an image of a target area and a preset driving route of the unmanned boat in the target area.

[0068] Specifically, the amphibious unmanned boat, as a means of transportation for patrol duty, emergency rescue and rapid transportation in the polar regions, can sail on various terrains such as ice and snow areas (ice layers and snow areas), water surfaces, etc. It can glide at high speed when sailing on the water surface, and can also travel smoothly under conditions such as ice and snow areas and drift ice areas. The image of the corresponding target area can be obtained through a satellite, or the image of the target area where the unmanned boat is located can be obtained through a panoramic vision system set above the unmanned boat. The image can be a panoramic view of a 360-degree perspective, or an image of a specified target area or direction. The image information of the target driving area of ​​the amphibious unmanned boat can be obtained through the image. The image may include different driving environments such as ice and snow areas or water areas. Therefore, when controlling the amphibious unmanned boat, it is first necessary to divide the image into regions to obtain at least one sub-region image, or it can also include multiple sub-region images at the same time. The sub-region division is divided according to the types of different regions actually contained in the image, and each sub-region image corresponds to a sub-region in the target area. For example, if the target area is snow, then the image obtained only contains snow, and the area type obtained is a snow area. However, if the target area includes both snow and ice, then the area divided by the corresponding image includes snow and ice areas, so that the corresponding driving strategy can be matched according to the driving environment corresponding to different areas. At the same time, the preset driving trajectory of the unmanned boat in the target area is obtained, and the specific area passed by the unmanned boat when driving in the target area can be determined through the preset driving route.

[0069] S200, determining abnormal pixels in the image according to a comparison result of the grayscale values ​​of each pixel in the image and corresponding adjacent pixels.

[0070] Specifically, the image is converted into a corresponding grayscale image, and the grayscale value corresponding to each pixel is obtained. Any pixel is taken as the central pixel, and the grayscale value of the central pixel is compared with the grayscale values ​​of the pixels of its adjacent image. Thus, the difference in grayscale values ​​between the central pixel and each adjacent pixel is used to determine whether the central pixel has a large difference with the adjacent pixels as a whole. If there is a large difference, the central pixel is determined as an abnormal pixel.

[0071] Exemplarily, a pixel point is randomly selected from the image as the central pixel point, and the 8 adjacent pixels around the central pixel point are taken as adjacent pixels. The difference between the central pixel point and each adjacent pixel point is determined by taking the difference between the pixel values ​​of the central pixel point and each adjacent pixel point. Since there may be obvious differences between the grayscale values ​​of pixels at the edge of the area in the image and the grayscale values ​​of the surrounding pixels, when the grayscale value difference between the central pixel point and the adjacent pixels is large, it can be preliminarily determined that the central pixel point may be a pixel point at the edge of the area in the image, that is, an abnormal pixel point. When the grayscale value difference between the central pixel point and the adjacent pixels is not large, it can be determined that the central pixel point is not at the edge of the image area, that is, it is determined as a non-abnormal pixel point.

[0072] It should be noted that the abnormal pixel points determined in this embodiment are not all located at the edge of the divided sub-area. Only the pixel points that may be located at the edge of the area are preliminarily screened. The preliminarily screened abnormal pixel points need to be further processed to obtain more accurate edge pixel points.

[0073] S300, selecting one of the abnormal pixel points as a reference point, and obtaining a deviation value corresponding to the selected abnormal pixel point according to a distance relationship between the reference point and each of the abnormal pixel points.

[0074] Specifically, a rectangular coordinate system is constructed according to the image, and the coordinates corresponding to each abnormal pixel point in the image are determined, and an abnormal pixel point is arbitrarily selected from all abnormal pixel points as a reference point, and the corresponding distance is calculated according to the coordinates of the reference point and each abnormal pixel point, and the distance between the reference point and itself is 0. All distances corresponding to the reference point are summed and normalized to obtain a deviation value corresponding to the currently selected abnormal pixel point, and all abnormal pixel points are traversed. The deviation value corresponding to each abnormal pixel point is calculated by the above method, and the degree of deviation of each abnormal pixel point from all abnormal pixel points can be judged by the deviation value.

[0075] Exemplarily, the deviation value satisfies:

[0076] ;

[0077] Among them, C k is the deviation value of the kth abnormal pixel, X k is the horizontal coordinate of the kth abnormal pixel point selected, Y k is the ordinate of the kth abnormal pixel point selected, X i is the horizontal coordinate of the ith abnormal pixel among all abnormal pixels, Y i is the ordinate of the i-th abnormal pixel among all abnormal pixels, and n is the number of all abnormal pixels.

[0078] S400, determining the abnormal pixel points whose deviation values ​​are less than a first preset threshold as edge pixel points, and dividing the image into regions according to all the edge pixel points to obtain at least one sub-region image, wherein each of the sub-region images corresponds to a sub-region of the target area.

[0079] Specifically, the positional relationship between each abnormal pixel and all abnormal pixels is determined by the deviation value of each abnormal pixel, that is, when the deviation value is large, it means that the abnormal pixel is relatively far away from all abnormal pixels, and when the deviation value is small, it means that the abnormal pixel is relatively close to all abnormal pixels. Edge pixels in the same area of ​​the image are all gathered near the edge of the area. Therefore, when the abnormal pixel is far away from all abnormal pixels, it means that the abnormal pixel may not be an edge pixel, so the pixel can be eliminated. By setting a reasonable first preset threshold, the abnormal pixels are screened according to the comparison result of the deviation value of each abnormal pixel and the first preset threshold to obtain edge pixels, that is, abnormal pixels with deviation values ​​less than the first preset threshold are determined as edge pixels, and abnormal pixels with deviation values ​​greater than or equal to the first preset threshold are determined as non-edge pixels.

[0080] Furthermore, the image of the target area usually includes a corresponding driving environment area, that is, a sub-area image corresponding to the sub-area in the target area. The sub-area may include different types of areas such as ice surface area, snow area and ice-water mixed area. Among them, the ice-water mixed area is mainly water, and the unmanned boat driving in it is equivalent to driving in water. There are also sub-area images corresponding to the sub-areas in the image of the target area, so that the image is divided according to the edge pixels of the area to obtain the corresponding sub-area images, and each sub-area image corresponds to a sub-area in the target area.

[0081] Exemplarily, the edge pixel points corresponding to different sub-regions are distributed at the edge positions of the corresponding regions in the image. Therefore, according to the distribution of all edge pixel points, it can be divided into at least one edge pixel point partition. This partition is larger than the final divided sub-region image. All edge pixel points in the partition are connected according to the set connection rules to obtain the corresponding curves. All the obtained curves are used to divide the image, and finally the divided sub-region image is obtained.

[0082] Specifically, an edge pixel point in the partition is selected as the first starting point, and the edge pixel point closest to it is connected, and then the connected edge pixel point is used as the second starting point, and the edge pixel point closest to the second starting point among the remaining edge pixels except the first starting point is found as the third starting point, and then the second starting point is connected to the third starting point, and so on, and finally a curve corresponding to the partition is generated, that is, a boundary curve, and the image of the area surrounded by the boundary curve is used as the divided sub-area image, such as Figure 2 The boundary curves obtained by the above method are smoothed to obtain the boundary curves corresponding to the sub-area images: boundary curve L1, boundary curve L2, boundary curve L3 and boundary curve L4, thereby dividing the image into: sub-area image 1, sub-area image 2, sub-area image 3, sub-area image 4 and sub-area image 5.

[0083] S500: Input the sub-region image into a trained image classification model, and output the region type corresponding to the sub-region image.

[0084] Specifically, images of different types of driving environments are obtained, and each driving environment image is labeled according to the type corresponding to the image, which is the region type. The region types include snow areas, ice areas, and ice-water mixed areas, etc. A model training set is constructed with all labeled driving environment images, and the initial classification model is iteratively trained through the model training set to obtain a trained image classification model, wherein the initial classification model may include a support vector machine, a multi-layer perceptron, or a deep learning model, etc. The trained image classification model outputs the region type corresponding to the sub-region image based on the input sub-region image, that is, the type corresponding to the driving environment in the sub-region image.

[0085] S600, determining the target sub-areas passed by the preset driving route when crossing the target area and the order of passing through the target sub-areas, and generating a control strategy for the unmanned boat according to the area type corresponding to the target sub-area and the order of passing through the target sub-areas.

[0086] Specifically, all sub-areas that the unmanned boat passes through when traveling along the preset driving route are determined according to the preset driving route, that is, all places that the unmanned boat passes through when crossing the target area, and the order of each sub-area passed, and the sub-area image corresponding to each sub-area passed is determined in the image, and the area type corresponding to the sub-area image is obtained, and the area type is used as the area type corresponding to the sub-area passed, and the area type corresponding to each sub-area passed is matched with the strategy of the unmanned boat driving in the area type, so as to obtain the driving strategy corresponding to each sub-area passed, and the corresponding driving strategies are sorted according to the order of all the sub-areas passed, and the control strategy of the unmanned boat crossing the target area along the preset driving route in the target area is generated. The unmanned boat is controlled to travel along the preset driving route by the control strategy. When it reaches the corresponding sub-area, the corresponding driving strategy in the area can be executed according to the generated control strategy, so that the unmanned boat can quickly and effectively respond to the changes in the driving environment in the driving route, and timely and accurately adjust and control the state according to the characteristics of the area type, thereby ensuring the adaptability of the control during the driving process of the unmanned boat, and avoiding the inability to adjust the driving state of the unmanned boat in time due to control lag, and ensuring the safety of the control of the unmanned boat.

[0087] For example, Figure 2As shown, the dotted line is the preset driving route, P1 is the starting point, P2 is the end point, and the order of the corresponding sub-region images in the image that the unmanned boat passes through along the preset driving route is: sub-region image 2-sub-region image 1-sub-region image 2-sub-region image 3-sub-region image 4, wherein the region type corresponding to sub-region image 1 is the ice surface region, the region type corresponding to sub-region image 2 is the snow region, the region type corresponding to sub-region image 3 is the ice-water mixed region, and the region type corresponding to sub-region image 4 is the snow region. As a result, the order of different region types that the unmanned boat passes through along the preset driving route is: snow region-ice surface region-snow region-ice-water mixed region-snow region. According to the characteristics of the unmanned boat driving in different region types, the driving strategy corresponding to each region type is pre-set. For example, the snow driving strategy: when driving on the snow, since the snow poses a greater resistance to the driving of the unmanned boat, the engine of the unmanned boat is controlled to output a greater power when driving in the snow area, thereby providing the unmanned boat with greater power and ensuring that the unmanned boat can drive quickly and stably in the snow; the ice driving strategy: when driving on the ice, since the friction between the ice surface and the unmanned boat is small, it is easy to cause side slip, and the raised ice blocks on the ice surface will also increase or decrease the difficulty of controlling the unmanned boat. Therefore, when driving on the ice, a more precise control strategy needs to be implemented, and the control data is optimized through fuzzy control, so that the unmanned boat can drive stably on the ice; the ice-water mixed area driving strategy, when driving in the ice-water mixed area, similar to driving in water, the unmanned boat will not be subject to excessive driving resistance and the risk of side slipping, so the engine output power can be appropriately reduced in the ice-water mixed area, thereby saving energy consumption and increasing the operating time of the unmanned boat. Finally, the control strategy for the preset driving route of the unmanned boat is obtained according to the order of area types: snow driving strategy-ice driving strategy-snow driving strategy-ice and water mixed area driving strategy-snow driving strategy, among which the execution time of each driving strategy is determined according to the location of the unmanned boat, so as to ensure that when the unmanned boat drives to the corresponding area, it can quickly adjust and execute the corresponding driving strategy.

[0088] In this embodiment, the pixel points in the image of the target area where the unmanned boat is traveling are screened, and the abnormal pixel points that may be located at the edge of the sub-area image in the image are obtained by comparing the grayscale values ​​of each pixel point. Then, an abnormal pixel point is selected, and the deviation value of the selected abnormal pixel point is determined by the distance relationship between the abnormal pixel point and all the abnormal pixel points, so that the deviation value corresponding to each selected abnormal pixel point can be obtained respectively, and the abnormal pixel points are further screened according to the comparison result of the deviation value of each abnormal pixel point and the first preset threshold value, and finally the edge pixel points corresponding to the sub-area image in the image are obtained. The edge position of the corresponding sub-area in the image can be more accurately judged by all the edge pixel points, so that the image can be accurately divided into regions according to all the edge pixel points to obtain the corresponding sub-area images, each sub-area image corresponds to a sub-area in the target area, and the sub-area images are classified by the trained image classification model to obtain the region type corresponding to the sub-area image, and the driving environment of the sub-area corresponding to the corresponding sub-area image can be obtained according to the region type, so that the appropriate driving strategy can be adjusted in time according to the region type, and the control stability and safety of the unmanned boat in different driving environments are improved. Through the preset driving route, the sub-areas that the unmanned boat passes through when crossing the target area and the order of passing the sub-areas are determined, so that the control strategy of the unmanned boat corresponding to the preset driving route can be generated according to the area type and order corresponding to the sub-areas passed. Through the driving strategy, the unmanned boat is controlled and adjusted accordingly according to different area types, avoiding the inability to adapt to different area types due to a single control method, improving the adjustment ability of the unmanned boat to cope with environmental changes during driving, and improving the adaptability of the unmanned boat control. It is also possible to make timely adjustments and controls to the unmanned boat when driving in different sub-areas to meet the control requirements of different types of sub-areas for the unmanned boat, effectively improving the safety of the unmanned boat control.

[0089] Optionally, determining abnormal pixels in the image according to a comparison result of the grayscale values ​​of each pixel in the image and corresponding adjacent pixels includes:

[0090] For a pixel point, determine the grayscale value difference between the pixel point and each of the adjacent pixel points to obtain the adjacent grayscale difference corresponding to each of the adjacent pixel points, wherein the adjacent pixel points are pixel points adjacent to the pixel point;

[0091] Determine the maximum adjacent grayscale difference, the minimum adjacent grayscale difference and the average adjacent grayscale difference according to all the adjacent grayscale differences;

[0092] Obtaining an abnormality score corresponding to the pixel point according to the maximum adjacent grayscale difference, the minimum adjacent grayscale difference and the average adjacent grayscale difference;

[0093] The pixel point whose abnormality score is greater than a second preset threshold is determined as the abnormal pixel point.

[0094] Optionally, the anomaly score satisfies:

[0095] ;

[0096] Where S is the abnormality score, X max is the maximum adjacent grayscale difference, X min is the minimum adjacent grayscale difference, is the average adjacent grayscale difference.

[0097] Exemplarily, the image is first converted into a grayscale image, and the grayscale value corresponding to each pixel therein is determined. A pixel is randomly selected as the central pixel and the 8 adjacent pixels adjacent to the central pixel, and the adjacent grayscale difference between the central pixel and each adjacent pixel is calculated respectively, that is, the central pixel is subtracted from each adjacent pixel and the absolute value is taken to obtain the adjacent grayscale difference corresponding to each adjacent pixel, thereby obtaining 8 adjacent grayscale differences corresponding to 8 adjacent pixels, and the 8 adjacent grayscale differences are added and averaged to obtain the average adjacent grayscale difference. At the same time, according to the values ​​corresponding to the 8 adjacent grayscale differences, the maximum adjacent grayscale difference with the largest value is selected. The grayscale difference and the minimum adjacent grayscale difference with the smallest value are obtained according to the product of the difference between the maximum adjacent grayscale difference and the average adjacent grayscale difference and the difference between the average adjacent grayscale difference and the minimum adjacent grayscale difference, to obtain the abnormality score corresponding to the central pixel, that is, the abnormality score corresponding to the selected pixel, so as to calculate the abnormality score corresponding to each pixel by the above method, and judge whether the pixel is an abnormal pixel according to the comparison result of the abnormality score corresponding to each pixel with the set second preset threshold, that is, when the abnormality score is greater than the second preset threshold, the corresponding pixel is determined as an abnormal pixel, otherwise, the corresponding pixel is a non-abnormal pixel.

[0098] In this optional embodiment, the abnormality score of the pixel is determined by the change in the grayscale value between each pixel and the surrounding adjacent pixels, so that the difference between the overall grayscale value of the pixel and all the surrounding adjacent pixels is judged through the abnormal score, and finally the abnormal pixels are screened by comparing the results with the second preset threshold, and the pixels with large differences preliminarily screened out are taken as abnormal pixels. The information of the boundary points of the area in the image can be obtained through the abnormal pixels, so that accurate area division can be performed according to the boundary point information, thereby improving the accuracy of image area division, and at the same time, the amount of subsequent data processing calculations is reduced by screening the abnormal pixels, thereby improving the efficiency of data processing.

[0099] Optionally, dividing the image into regions according to all the edge pixels to obtain at least one sub-region image includes:

[0100] According to all the edge pixels, a boundary curve is obtained by curve fitting;

[0101] The image is divided according to the boundary curve to obtain the sub-region image.

[0102] Optionally, obtaining a boundary curve by curve fitting according to all the edge pixels includes:

[0103] Generating at least one edge pixel point set according to the distribution of all the edge pixel points;

[0104] According to the edge pixel point set, a plurality of fitting curves are generated by a preset fitting model;

[0105] The fitting curve with the smallest deviation from the edge pixel point set is determined as the boundary curve.

[0106] Specifically, based on the preset fitting model, at least one fitting curve is generated according to the obtained edge pixel points. If there is only one type of region in the image, the final fitting curve obtained by fitting may be the edge curve of the image. For example, when there is only snow in the target area, the edge pixel points obtained are the pixel points at the edge of the image, and the final fitting curve obtained is the edge curve of the image edge, and the final sub-region image obtained is the image itself. When there are multiple region types in the target area, multiple corresponding boundary curves will eventually be obtained, and the image will be divided into multiple sub-region images. Since the region in the image may not be a complete closed area, the number of multiple boundary curves here may be greater than the number of sub-region images. The fitting model includes a preset multi-order polynomial. According to the distribution of edge pixels, the edge pixels are divided into at least one edge pixel set. The edge pixel set includes edge pixels of corresponding areas in the image. If there are multiple concentrated areas of edge pixel distribution in the image, multiple edge pixel sets are generated accordingly. All edge pixels in the edge pixel set are fitted and calculated by the preset multi-order polynomial. The coefficients in the multi-order polynomial are continuously adjusted. A corresponding fitting curve is generated according to each combination of coefficients. The deviation between each fitting curve and the overall deviation of all edge pixels in the corresponding edge pixel set is calculated respectively. The fitting curve with the smallest deviation is selected as the final boundary curve of the edge pixel set. When there are multiple edge pixel sets, multi-order polynomials that conform to the respective edge pixel sets can be generated by the fitting model to obtain the boundary curve corresponding to each edge pixel set. Finally, the image is divided into sub-area images according to all the generated boundary curves.

[0107] For example, Figure 2 As shown, by fitting the edge pixels in the image, boundary curves L1, L2, L3 and L4 are obtained, and the image is divided into regions according to the four boundary curves, wherein the edges of the image are also used as boundary lines for region division, and the sub-region images finally obtained are: sub-region image 1, sub-region image 2, sub-region image 3, sub-region image 4 and sub-region image 5. To ensure that each sub-region image does not overlap, sub-region image 2 is the remaining image after sub-region image 1 is removed.

[0108] In this optional embodiment, the boundary curve of the image is obtained by fitting the edge pixels, and the image is divided by the boundary curve to obtain a sub-region image corresponding to the sub-region in the target region, and then the corresponding region type can be obtained according to the classification result of the sub-region image, and a suitable driving strategy can be matched for the corresponding sub-region according to the region type, so that the unmanned boat can be controlled to drive quickly and stably in the corresponding area through the driving strategy, avoiding the inability to adjust the suitable control strategy in time due to the single control method, effectively improving the adaptability of the unmanned boat to different driving environments during the control process, and ensuring the safe and stable driving of the unmanned boat.

[0109] Optionally, inputting the sub-region image into a trained image classification model and outputting the region type corresponding to the sub-region image includes:

[0110] Determine the number of pixels corresponding to each grayscale value according to the grayscale value of each pixel in the sub-region image;

[0111] Determine the grayscale value ratio corresponding to each grayscale value according to the ratio of the number of pixels corresponding to each grayscale value to the number of all pixels;

[0112] Adding the product of each grayscale value and the corresponding grayscale value ratio to obtain a complementary grayscale value;

[0113] Completing the sub-region image according to the completed grayscale value to generate a sub-region image of a preset size;

[0114] The sub-region image of the preset size is input into the image classification model to obtain the corresponding region type.

[0115] Specifically, since the divided sub-region image is not a complete regular image, the sub-region image needs to be complemented to ensure that the image classification model recognizes an image of the same size, thereby improving the processing speed and accuracy of the image. First, according to the grayscale value of each pixel in the sub-region image, the number of pixels with the same grayscale value is counted, that is, the number of pixels corresponding to each grayscale value in the sub-region image, and the grayscale value ratio corresponding to the grayscale value is determined according to the ratio of the number of pixels of the grayscale value to the number of all pixels in the sub-region image. The grayscale value ratio corresponding to each grayscale value in the sub-region image is calculated by the above method, and the product of each grayscale value and the corresponding grayscale value ratio is added to obtain the complement grayscale value corresponding to the sub-region image. For example, there are 100 pixels in the sub-region image, of which 50 types of different grayscale values ​​are included. Then, the number of pixels corresponding to the 50 grayscale values ​​is counted, and the number of pixels corresponding to each grayscale value is divided by 100 to obtain the grayscale value ratio of each grayscale value. Finally, the product of each grayscale value and the corresponding grayscale value ratio is added to obtain the complement grayscale value. The pixels that need to be completed in the sub-region image are determined according to the preset size, and the completed grayscale value is set to the grayscale value of the completed pixel. Finally, the completed sub-region image is input into the image classification model, and the region type corresponding to the sub-region image is output.

[0116] In this optional embodiment, the sub-region image is completed by completing the grayscale value, which ensures the integrity of the sub-region image information, avoids the influence of vacant positions on the classification of the sub-region image, and improves the accuracy of the classification of the sub-region image, so as to generate the corresponding driving strategy according to the accurate region classification result, and at the same time, by adding the product of each grayscale value and the corresponding grayscale value ratio, the completed grayscale value suitable for the grayscale value and the overall grayscale of the sub-region image is obtained. By completing with the completed grayscale value, it is possible to avoid the grayscale value of the completed pixel point interfering with the sub-region image, thereby ensuring the accuracy of region classification according to the completed sub-region image.

[0117] Optionally, the determining the target sub-areas that the preset driving route passes through when crossing the target area and the order of passing through the target sub-areas includes:

[0118] Determine the route pixel points corresponding to the preset driving route in the image;

[0119] Determine a driving sub-region image from among all the sub-region images according to a comparison result between the route pixel points and each of the sub-region images, and determine a sub-region corresponding to the driving sub-region image as the target sub-region;

[0120] The image sequence of the corresponding driving sub-area images is determined according to the sequence of the route pixels in the preset driving route, wherein the image sequence is the sequence in which the preset driving route passes through the target sub-area.

[0121] Optionally, determining the driving sub-region image in the sub-region image according to the comparison result between the route pixel point and each of the sub-region images includes:

[0122] When one of the sub-region images coincides with a pixel point of the route, the sub-region image is determined as the driving sub-region image;

[0123] When a sub-region image does not overlap with a pixel point of the route, the sub-region image is determined as a non-driving sub-region image.

[0124] Specifically, Figure 2 As shown, the dotted line is the preset driving route, P1 is the starting point, and P2 is the end point. The corresponding route pixel points in the image are determined according to the preset driving route, that is, Figure 2 The pixel points corresponding to the dotted line in the middle are compared with the position of each sub-region image by comparing the route pixel points. When a sub-region image coincides with the route pixel points, it is determined that the preset driving route passes through the sub-region image, and the sub-region image is determined as the driving sub-region image. When a sub-region image does not coincide with the route pixel points, it is determined that the preset driving route does not pass through the sub-region image, and the sub-region image is determined as the non-driving sub-region image. Thus, all the driving sub-region images that the preset driving route passes through in the image, namely, sub-region image P1, sub-region image P2, sub-region image P3 and sub-region image P4, are obtained by comparison. Figure 2 It can be seen that the preset driving route does not pass through sub-area image 5. When the unmanned boat starts from the starting point P1 and travels along the preset driving route to the end point P2, the image order of the driving sub-area images passed is: sub-area image 2-sub-area image 1-sub-area image 2-sub-area image 3-sub-area image 4, and the sub-area corresponding to each driving sub-area image in the target area is determined as the target sub-area, so that the order of the target sub-areas mentioned in the corresponding preset driving route is obtained according to the image order of the driving area images, wherein the target sub-area is the sub-area that the unmanned boat actually passes through during the driving process in the target area.

[0125] It should be noted that the preset driving route can be adjusted according to actual conditions. For example, when the preset driving route passes through the junction of the ice surface area and the ice-water mixed area, due to the difference between the ice surface area and the ice-water mixed area, that is, the difference between the buoyancy of the ice surface supporting the hull and the buoyancy of the ice-water mixed area on the hull, if the unmanned boat is driving at this junction, it may be subject to uneven buoyancy changes, causing the unmanned boat to tilt and capsize. Therefore, the driving route can be adjusted to only pass through the ice surface area or only pass through the ice-water mixed area. Adaptive adjustments can be made according to the actual driving status. Usually, the driving route is adjusted to pass through the ice-water mixed area. Since it is easier to control in the water, the complexity of control can be reduced.

[0126] In this optional embodiment, by presetting the driving route and the corresponding positional relationship between each sub-region image on the driving route, the driving sub-region images passed by the unmanned boat during the driving process, as well as the order of passing each driving sub-region image, are obtained, and the target sub-regions in the target area are determined according to the sub-regions in the target area corresponding to the driving sub-region images, that is, the sub-regions passed by the unmanned boat when driving in the target area, and the order of passing the target sub-regions. Then, according to the target sub-regions and the corresponding order, the change of different area types and the change order of the unmanned boat in the target area can be obtained, so that the corresponding driving strategy can be matched according to the change order, so that the unmanned boat will make corresponding control adjustments when it drives to each target sub-region, which improves the ability of the unmanned boat to cope with changes in the driving environment during the driving process, and effectively improves the adaptability of the control of the unmanned boat.

[0127] Optionally, generating the control strategy of the unmanned boat according to the area type corresponding to the target sub-area and the order of passing through the target sub-area includes:

[0128] Determining a driving strategy corresponding to the target sub-area according to the area type corresponding to the target sub-area and a preset matching rule, wherein the matching rule includes a preset one-to-one correspondence between the area type and the driving strategy;

[0129] The determined driving strategies are sorted according to the order in which the preset driving routes pass through the target sub-areas to generate the control strategy of the unmanned boat.

[0130] Specifically, corresponding driving strategies are set according to different area types, that is, suitable control and adjustment schemes are executed for different driving environments, so as to ensure that the unmanned boat can operate quickly and stably in the corresponding driving environment, and matching rules are generated through area types and corresponding driving strategies, that is, a one-to-one correspondence between area types and driving strategies. According to the area type corresponding to each target sub-area, the corresponding driving strategy is matched in the matching rule, and the order of the corresponding driving strategies is determined according to the order of passing the target sub-areas, so as to generate a control strategy for the unmanned boat to travel along the preset driving route according to all driving strategies and the corresponding order.

[0131] For example, Figure 2 The image sequence of the driving sub-area images passed by the unmanned boat is: sub-area image 2-sub-area image 1-sub-area image 2-sub-area image 3-sub-area image 4. Since the sub-area images in the image correspond to the sub-areas in the target area one-to-one, the image sequence is the sequence in which the unmanned boat drives through the corresponding target sub-areas in the target area, wherein the area type of sub-area image 1 is an ice surface area, the area type of sub-area image 2 is a snow area, the area type of sub-area image 3 is an ice-water mixed area, and the area type of sub-area image 4 is a snow area. Therefore, the area type sequence corresponding to the unmanned boat passing through the target area is: snow area-ice surface area-snow area-ice-water mixed area-snow area. The preset matching rule is: snow area corresponds to snow driving strategy, ice surface area corresponds to ice driving strategy, and ice-water mixed area corresponds to ice-water mixed area driving strategy. Therefore, the final control strategy obtained according to the order of area types and the matching rule is: snow driving strategy-ice surface driving strategy-snow driving strategy-ice-water mixed area driving strategy-snow driving strategy. It should be noted that the length of the driving route in the driving sub-area image can be determined based on the number of route pixel points in the driving sub-area image, and the driving time of the unmanned boat in different driving group areas and the time point for switching different driving strategies can be determined based on the driving length and the corresponding driving speed.

[0132] In this optional embodiment, the corresponding driving strategy is matched according to the area type of each driving sub-area image through matching rules. The matched driving strategy can ensure that the unmanned boat can be quickly and accurately adjusted and controlled after entering the corresponding area, thereby improving the adaptability of the unmanned boat to different driving environments during driving. The driving strategies are sorted according to the order in which the preset route passes through different areas, and the control strategy of the unmanned boat is generated. The control strategy can make rapid and effective adjustments to the regional changes during the driving process of the unmanned boat, avoiding fluctuations in the driving route of the unmanned boat or even deviation from the route due to adjustment lags, thereby effectively improving the safety of the control of the unmanned boat.

[0133] like Figure 3As shown, an embodiment of the present invention provides a sensing system 300 for an amphibious unmanned boat, including:

[0134] An acquisition module 310 is used to acquire an image of a target area and a preset driving route of the unmanned boat in the target area;

[0135] A screening module 320, configured to determine abnormal pixels in the image based on a comparison result of the grayscale values ​​of each pixel in the image and corresponding adjacent pixels;

[0136] The processing module 330 is used to select one of the abnormal pixel points as a reference point, and obtain a deviation value corresponding to the selected abnormal pixel point according to the distance relationship between the reference point and each of the abnormal pixel points;

[0137] A comparison module 340 is used to determine the abnormal pixel points whose deviation values ​​are less than a first preset threshold as edge pixel points, and divide the image into regions according to all the edge pixel points to obtain at least one sub-region image, wherein each of the sub-region images corresponds to a sub-region of the target region;

[0138] A classification module 350, configured to input the sub-region image into a trained image classification model and output a region type corresponding to the sub-region image;

[0139] The generation module 360 ​​is used to determine the target sub-areas passed by the preset driving route when crossing the target area and the order of passing through the target sub-areas, and generate the control strategy of the unmanned boat according to the area type corresponding to the target sub-area and the order of passing through the target sub-areas.

[0140] The perception system for amphibious unmanned boats of this embodiment is used to implement the perception method for amphibious unmanned boats as described above. Its advantages over the prior art are the same as the advantages of the perception method for amphibious unmanned boats as described above, and will not be repeated here.

[0141] like Figure 4 As shown, an electronic device 400 provided by an embodiment of the present invention includes a memory 410 and a processor 420; the memory 410 is used to store a computer program; the processor 420 is used to implement the above-mentioned perception method for amphibious unmanned boats when executing the computer program.

[0142] In other words, an electronic device 400 includes a memory 410 and a processor 420 coupled to the memory 410; the memory 410 is configured to store a computer program; and the processor 420 is configured to perform the following operations when executing the computer program:

[0143] Acquire an image of a target area and a preset driving route of the unmanned boat in the target area;

[0144] Determine abnormal pixels in the image based on a comparison result of the grayscale values ​​of each pixel in the image and corresponding adjacent pixels;

[0145] Selecting one of the abnormal pixel points as a reference point, and obtaining a deviation value corresponding to the selected abnormal pixel point according to a distance relationship between the reference point and each of the abnormal pixel points;

[0146] Determine the abnormal pixel points whose deviation values ​​are less than a first preset threshold as edge pixel points, and divide the image into regions according to all the edge pixel points to obtain at least one sub-region image, wherein each sub-region image corresponds to a sub-region of the target region;

[0147] Inputting the sub-region image into a trained image classification model, and outputting the region type corresponding to the sub-region image;

[0148] Determine the target sub-areas that the preset driving route passes through when crossing the target area and the order of passing through the target sub-areas, and generate a control strategy for the unmanned boat according to the area type corresponding to the target sub-area and the order of passing through the target sub-areas.

[0149] An embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the perception method for amphibious unmanned boats as described above is implemented.

[0150] In other words, a non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor performs the following operations:

[0151] Acquire an image of a target area and a preset driving route of the unmanned boat in the target area;

[0152] Determine abnormal pixels in the image based on a comparison result of the grayscale values ​​of each pixel in the image and corresponding adjacent pixels;

[0153] Selecting one of the abnormal pixel points as a reference point, and obtaining a deviation value corresponding to the selected abnormal pixel point according to a distance relationship between the reference point and each of the abnormal pixel points;

[0154] Determine the abnormal pixel points whose deviation values ​​are less than a first preset threshold as edge pixel points, and divide the image into regions according to all the edge pixel points to obtain at least one sub-region image, wherein each sub-region image corresponds to a sub-region of the target region;

[0155] Inputting the sub-region image into a trained image classification model, and outputting the region type corresponding to the sub-region image;

[0156] Determine the target sub-areas that the preset driving route passes through when crossing the target area and the order of passing through the target sub-areas, and generate a control strategy for the unmanned boat according to the area type corresponding to the target sub-area and the order of passing through the target sub-areas.

[0157] Now will describe the electronic device 400 that can be used as the server or client of the present invention, it is the example of the hardware device that can be applied to each aspect of the present invention.Electronic device 400 is intended to represent the computer device of various digital electronics of travel, such as, laptop computer, desktop computer, workbench, personal digital assistant, server, blade server, mainframe computer and other suitable computer.Electronic device 400 can also represent various mobile devices of travel, such as, personal digital processing, cellular phone, smart phone, wearable device and other similar computing devices.The components shown in this article, their connection and relationship, and their functions are only as examples, and are not intended to limit the realization of the present invention described and / or required herein.

[0158] The electronic device 400 includes a computing unit, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) or a computer program loaded from a storage unit into a random access memory (RAM). In the RAM, various programs and data required for the operation of the device can also be stored. The computing unit, ROM, and RAM are connected to each other via a bus. An input / output (I / O) interface is also connected to the bus.

[0159] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc. In the present application, the unit described as a separate component may or may not be physically separated, and the component displayed as a unit may or may not be a physical unit, that is, it may be located in one place, or it may be distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment of the present invention. In addition, each functional unit in each embodiment of the present invention can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional unit.

[0160] Although the present invention is disclosed as above, the protection scope of the present invention is not limited thereto. Those skilled in the art may make various changes and modifications without departing from the spirit and scope of the present invention, and these changes and modifications will fall within the protection scope of the present invention.

Claims

1. A perception method for amphibious unmanned boats, characterized in that: include: Acquire an image of a target area and a preset driving route of the unmanned boat in the target area; Determine abnormal pixels in the image based on a comparison result of the grayscale values ​​of each pixel in the image and corresponding adjacent pixels; Select one of the abnormal pixel points as a reference point, and obtain the deviation value corresponding to the selected abnormal pixel point according to the distance relationship between the reference point and each abnormal pixel point, wherein the corresponding distance is calculated according to the coordinates of the reference point and each abnormal pixel point, and the distance between the reference point and itself is 0. All the distances corresponding to the reference point are summed and normalized to obtain the deviation value corresponding to the currently selected abnormal pixel point, and all abnormal pixels are traversed; Determine the abnormal pixel points whose deviation values ​​are less than a first preset threshold as edge pixel points, and divide the image into regions according to all the edge pixel points to obtain at least one sub-region image, wherein each sub-region image corresponds to a sub-region of the target region; Inputting the sub-region image into a trained image classification model, and outputting the region type corresponding to the sub-region image; Determine the target sub-areas that the preset driving route passes through when crossing the target area and the order of passing through the target sub-areas, and generate a control strategy for the unmanned boat according to the area type corresponding to the target sub-area and the order of passing through the target sub-areas.

2. The sensing method for amphibious unmanned boat according to claim 1, characterized in that: The step of determining abnormal pixels in the image according to a comparison result of the grayscale values ​​of each pixel in the image and corresponding adjacent pixels includes: For a pixel point, determine the grayscale value difference between the pixel point and each of the adjacent pixel points to obtain the adjacent grayscale difference corresponding to each of the adjacent pixel points, wherein the adjacent pixel points are pixel points adjacent to the pixel point; Determine the maximum adjacent grayscale difference, the minimum adjacent grayscale difference and the average adjacent grayscale difference according to all the adjacent grayscale differences; Obtaining an abnormality score corresponding to the pixel point according to the maximum adjacent grayscale difference, the minimum adjacent grayscale difference and the average adjacent grayscale difference; The pixel point whose abnormality score is greater than a second preset threshold is determined as the abnormal pixel point.

3. The sensing method for amphibious unmanned boat according to claim 2 is characterized in that: The anomaly score satisfies: ； Where S is the abnormality score, X max is the maximum adjacent grayscale difference, X min is the minimum adjacent grayscale difference, is the average adjacent grayscale difference.

4. The sensing method for amphibious unmanned boat according to claim 1, characterized in that: The step of dividing the image into regions according to all the edge pixels to obtain at least one sub-region image comprises: According to all the edge pixels, a boundary curve is obtained by curve fitting; The image is divided according to the boundary curve to obtain the sub-region image.

5. The sensing method for amphibious unmanned boat according to claim 4 is characterized in that: The step of obtaining a boundary curve by curve fitting based on all the edge pixels includes: Generating at least one edge pixel point set according to the distribution of all the edge pixel points; According to the edge pixel point set, a plurality of fitting curves are generated by a preset fitting model; The fitting curve with the smallest deviation from the edge pixel point set is determined as the boundary curve.

6. The sensing method for amphibious unmanned boat according to claim 1, characterized in that: The step of inputting the sub-region image into a trained image classification model and outputting the region type corresponding to the sub-region image comprises: Determine the number of pixels corresponding to each grayscale value according to the grayscale value of each pixel in the sub-region image; Determine the grayscale value ratio corresponding to each grayscale value according to the ratio of the number of pixels corresponding to each grayscale value to the number of all pixels; Adding the product of each grayscale value and the corresponding grayscale value ratio to obtain a complementary grayscale value; Completing the sub-region image according to the completed grayscale value to generate a sub-region image of a preset size; The sub-region image of the preset size is input into the image classification model to obtain the corresponding region type.

7. The sensing method for amphibious unmanned boat according to claim 1, characterized in that: The determining the target sub-areas that the preset driving route passes through when crossing the target area and the order of passing through the target sub-areas includes: Determine the route pixel points corresponding to the preset driving route in the image; Determine a driving sub-region image from among all the sub-region images according to a comparison result between the route pixel points and each of the sub-region images, and determine a sub-region corresponding to the driving sub-region image as the target sub-region; The image sequence of the corresponding driving sub-area images is determined according to the sequence of the route pixels in the preset driving route, wherein the image sequence is the sequence in which the preset driving route passes through the target sub-area.

8. The sensing method for amphibious unmanned boat according to claim 7, characterized in that: The step of determining the driving sub-region image from among all the sub-region images according to the comparison result between the route pixel points and each of the sub-region images comprises: When one of the sub-region images coincides with a pixel point of the route, the sub-region image is determined as the driving sub-region image; When a sub-region image does not overlap with a pixel point of the route, the sub-region image is determined as a non-driving sub-region image.

9. The sensing method for amphibious unmanned boat according to claim 1, characterized in that: The step of generating the control strategy of the unmanned boat according to the area type corresponding to the target sub-area and the order of passing through the target sub-area includes: Determining a driving strategy corresponding to the target sub-area according to the area type corresponding to the target sub-area and a preset matching rule, wherein the matching rule includes a preset one-to-one correspondence between the area type and the driving strategy; The determined driving strategies are sorted according to the order in which the preset driving routes pass through the target sub-areas to generate the control strategy of the unmanned boat.

10. A sensing system for an amphibious unmanned boat, characterized in that: include: An acquisition module, used to acquire an image of a target area and a preset driving route of the unmanned boat in the target area; A screening module, used for determining abnormal pixels in the image based on a comparison result of the grayscale values ​​of each pixel in the image and the corresponding adjacent pixels; A processing module is used to select one of the abnormal pixel points as a reference point, and obtain a deviation value corresponding to the selected abnormal pixel point according to the distance relationship between the reference point and each abnormal pixel point, wherein the corresponding distance is calculated according to the coordinates of the reference point and each abnormal pixel point, the distance between the reference point and itself is 0, and all the distances corresponding to the reference point are summed and normalized to obtain the deviation value corresponding to the currently selected abnormal pixel point, and all abnormal pixels are traversed; A comparison module, configured to determine the abnormal pixel points whose deviation value is less than a first preset threshold as edge pixel points, and divide the image into regions according to all the edge pixel points to obtain at least one sub-region image, wherein each of the sub-region images corresponds to a sub-region of the target region; A classification module, used for inputting the sub-region image into a trained image classification model and outputting the region type corresponding to the sub-region image; A generation module is used to determine the target sub-areas passed by the preset driving route when crossing the target area and the order of passing through the target sub-areas, and generate a control strategy for the unmanned boat according to the area type corresponding to the target sub-area and the order of passing through the target sub-areas.

Images