CN116543398B - Ship name and violation identification method and system - Google Patents

Abstract

The invention discloses a ship name and violation identification method and system. The method introduces laser, and performs fixed-point light filling at night; processing a multi-vessel scenario by the relative positions of vessels in the panoramic camera and the close-up camera; the artificial intelligence is used for respectively identifying the elements such as the bow, the hull, the stern, the ship plate, the characters and the like so as to avoid the influence of the background as much as possible; introducing alternative AIS data and fuzzy query of a database to defects and shielding, comprehensively analyzing factors such as the occurrence frequency of characters to calculate candidate matching degree and digital matching degree, and judging whether a ship name is sleeved, whether the ship name is not registered and whether the AIS is not opened according to the matching degree score; and finally determining whether the actual candidate recognition ship name accords with the matched result ship name or not through the deviation distance of the candidate recognition ship name and the matched result ship name. The method can accurately identify the ship name and judge the condition of violation.

Description

Ship name and violation identification method and system

Technical Field

The invention relates to the technical field of ships, in particular to a ship name and violation identification method and system.

Background

In a common system, the ship name recognition is usually carried out only by artificial intelligence, and is influenced by factors such as insufficient illumination at night, indistinguishable multi-ship intersection, large background interference (shop signboards and banners in the background, other characters printed on a ship-borne container), pollution damage or shielding interference (the ship name is shielded by hanging tires and the like), and the recognition rate is low, so that the method is difficult to be truly applied to the illegal recognition. In the traditional method, because the identified ship name content cannot be accurately judged, database inquiry is generally introduced to complement identification, and the requirement of a user on the actual integrity inspection of the ship name content cannot be effectively met.

Aiming at the problem that the ship name content cannot be accurately judged and identified in the prior art, no effective solution is proposed at present.

Disclosure of Invention

The embodiment of the invention provides a ship name and violation identification method and system, which are used for solving the problem that the content of the ship name cannot be accurately judged and identified in the prior art.

In order to achieve the above object, in one aspect, the present invention provides a method for identifying a ship name and a violation, the method comprising: s1, acquiring a panoramic image and a plurality of close-up images of a current ship through a laser identification system, a panoramic camera and a close-up camera; acquiring AIS data of all ships within a specified radius range through an AIS base station, and taking the AIS data as candidate AIS data; s2, judging the ship category according to the panoramic image; performing ship plate recognition on the close-up images, and deleting the invalid images to obtain a plurality of effective ship plate images; s3, when the effective ship plate image is judged to be empty, alarming if the ship name is not recognized; otherwise, performing character recognition on the plurality of effective ship board images to obtain a plurality of recognition candidates and confidence degrees corresponding to the recognition candidates; generating a plurality of candidate results according to the identified candidates; performing fuzzy processing and reverse order processing on the candidate result to obtain a plurality of candidate identifications and confidence degrees corresponding to each candidate identification; s4, when all the confidence degrees of the candidate identification are judged not to exceed a first preset threshold value, alarming by the fact that the ship name is not identified; otherwise, sequencing all candidate identifications from high to low according to the confidence level and matching with the candidate AIS data; s5, when the alternative AIS data are judged to be empty, an AIS alarm is not opened, and a step S6 is carried out; otherwise, calculating the candidate matching degree and the digital matching degree of each candidate identification and the candidate AIS data; when judging that the candidate matching degree does not exceed the second preset threshold value or the digital matching degree does not exceed the third preset threshold value, considering that the current ship is not opened with AIS or is covered with cards, and entering step S6; otherwise, the candidate AIS data with the highest candidate matching degree is taken as matching AIS data; judging whether the matched AIS data are matched or not, if so, warning by fake plate; and proceeds to step S7; s6, sequencing all candidate identifications from high to low according to the confidence level, matching the candidate identifications with ship names in the database, and calculating the candidate matching degree and the digital matching degree of each candidate identification and the ship name of the database; when the digital matching degree is judged not to exceed a third preset threshold value, warning by a non-registered ship name; otherwise, taking the database ship name with the highest candidate matching degree as a matching ship name, and entering step S7; and S7, judging whether the matching degree between the matching AIS data or the matching ship name and the corresponding candidate identification is 100%, if so, obtaining the ship name of the ship, otherwise, judging whether the ship board is blocked or stained according to the deviation distance between the matching AIS data or the matching ship name and the corresponding candidate identification.

Optionally, the invalid image includes: at least one of the bow, the hull and the stern, otherwise, the image is invalid; the image excluding the ship board is regarded as an invalid image; at least one of four corners of the identification frame of the ship plate is arranged in the bow or the hull or the stern, and the center point of the identification frame of the ship plate is arranged in the bow or the hull or the stern, otherwise, the identification frame of the ship plate is regarded as an invalid image; the relative offset of the center pixel coordinate of the identification frame of the ship plate from the center pixel coordinate of the close-up image is larger than a fourth preset threshold value, and the ship plate is regarded as an invalid image; the maximum length of the ship plate is smaller than the fifth preset threshold value and is regarded as an invalid image.

Optionally, generating a plurality of candidate results according to the identified candidates; the fuzzy processing and the reverse order processing are carried out on the candidate result, and the obtaining of a plurality of candidate identifications and the confidence degree corresponding to each candidate identification comprises the following steps: combining all words in the recognition candidates to generate a plurality of candidate results; adding wildcards into all candidate results to generate fuzzy results so as to carry out fuzzy matching; performing reverse order processing on all the fuzzy results to obtain reverse order results; the fuzzy result and the reverse order result are identified as a plurality of candidates.

Optionally, when the candidate matching degree is determined not to exceed the second preset threshold or the digital matching degree is determined not to exceed the third preset threshold, the step of considering that the current ship does not open the AIS or the quilt cover includes: after all ships leave the specified radius range, if the rest unmatched alternative AIS data with the same direction exist, the current ship is considered to be sleeved; if no unmatched alternative AIS data remain, an AIS alarm is not opened.

Optionally, when the number matching degree is not greater than the third preset threshold, the alarming with the unregistered ship name includes: and when the number matching degree is judged not to exceed a third preset threshold value, adjusting the default of each character in each candidate identification to match with the ship name in the database, and if the matching is still unsuccessful, alarming by using the unregistered ship name.

Optionally, the S1 includes: the position of the ship is identified through the laser identification system, and a plurality of close-up images are continuously shot through the close-up camera after the ship starts to light at a preset distance before traveling to a preset range; and aligning the central axes of the laser identification system and the panoramic camera so as to enable the ship to travel to the laser prevention area, continuously shooting a plurality of images through the panoramic camera, and taking the image in the middle of shooting time as a panoramic image.

In another aspect, the present invention provides a ship name and violation identification system, comprising: the acquisition unit is used for acquiring the panoramic image and a plurality of close-up images of the current ship through the laser identification system, the panoramic camera and the close-up camera; acquiring AIS data of all ships within a specified radius range through an AIS base station, and taking the AIS data as candidate AIS data; a deleting unit for judging the ship category according to the panoramic image; performing ship plate recognition on the close-up images, and deleting the invalid images to obtain a plurality of effective ship plate images; the first judging unit is used for alarming when the effective ship plate image is judged to be empty and the ship name is not recognized; otherwise, performing character recognition on the plurality of effective ship board images to obtain a plurality of recognition candidates and confidence degrees corresponding to the recognition candidates; generating a plurality of candidate results according to the identified candidates; performing fuzzy processing and reverse order processing on the candidate result to obtain a plurality of candidate identifications and confidence degrees corresponding to each candidate identification; the second judging unit is used for alarming when the confidence level of all the candidate identifications is judged not to exceed the first preset threshold value; otherwise, sequencing all candidate identifications from high to low according to the confidence level and matching with the candidate AIS data; the third judging unit is used for alarming when the alternative AIS data is judged to be empty and entering the database matching unit when the AIS data is not opened; otherwise, calculating the candidate matching degree and the digital matching degree of each candidate identification and the candidate AIS data; when the candidate matching degree is judged not to exceed the second preset threshold value or the digital matching degree is judged not to exceed the third preset threshold value, the current ship is considered to be not opened with AIS or is not covered with cards, and the ship is entered into a database matching unit; otherwise, the candidate AIS data with the highest candidate matching degree is taken as matching AIS data; judging whether the matched AIS data are matched or not, if so, warning by fake plate; and then enters a matching degree calculating unit; the database matching unit is used for sequencing all candidate identifications from high to low according to the confidence level, matching the candidate identifications with the ship names in the database, and calculating the candidate matching degree and the digital matching degree of each candidate identification and the ship name of the database; when the digital matching degree is judged not to exceed a third preset threshold value, warning by a non-registered ship name; otherwise, taking the database ship name with the highest candidate matching degree as a matching ship name, and entering a matching degree calculation unit; and the matching degree calculation unit is used for judging whether the matching degree between the matching AIS data or the matching ship name and the corresponding candidate identification is 100%, if so, obtaining the ship name of the ship, otherwise, judging whether the ship board is blocked or stained according to the deviation distance between the matching AIS data or the matching ship name and the corresponding candidate identification.

Optionally, generating a plurality of candidate results according to the identified candidates; the fuzzy processing and the reverse order processing are carried out on the candidate result, and the obtaining of a plurality of candidate identifications and the confidence degree corresponding to each candidate identification comprises the following steps: combining all words in the recognition candidates to generate a plurality of candidate results; adding wildcards into all candidate results to generate fuzzy results so as to carry out fuzzy matching; performing reverse order processing on all the fuzzy results to obtain reverse order results; the fuzzy result and the reverse order result are identified as a plurality of candidates.

Optionally, when the candidate matching degree is determined not to exceed the second preset threshold or the digital matching degree is determined not to exceed the third preset threshold, the step of considering that the current ship does not open the AIS or the quilt cover includes: after all ships leave the specified radius range, if the rest unmatched alternative AIS data with the same direction exist, the current ship is considered to be sleeved; if no unmatched alternative AIS data remain, an AIS alarm is not opened.

Optionally, when the number matching degree is not greater than the third preset threshold, the alarming with the unregistered ship name includes: and when the number matching degree is judged not to exceed a third preset threshold value, adjusting the default of each character in each candidate identification to match with the ship name in the database, and if the matching is still unsuccessful, alarming by using the unregistered ship name.

The invention has the beneficial effects that:

the invention provides a ship name and violation identification method and system, wherein laser is introduced into the method, and fixed-point light filling is carried out at night; processing a multi-vessel scenario by the relative positions of vessels in the panoramic camera and the close-up camera; the artificial intelligence is used for respectively identifying the elements such as the bow, the hull, the stern, the ship plate, the characters and the like so as to avoid the influence of the background as much as possible; introducing alternative AIS data and fuzzy query of a database to defects and shielding, comprehensively analyzing factors such as the occurrence frequency of characters to calculate candidate matching degree and digital matching degree, and judging whether a ship name is sleeved, whether the ship name is not registered and whether the AIS is not opened according to the matching degree score; and finally determining whether the actual candidate recognition ship name accords with the matched result ship name or not through the deviation distance of the candidate recognition ship name and the matched result ship name. The method can accurately identify the ship name and judge the condition of violation.

Drawings

FIG. 1 is a flow chart of a method for identifying ship names and violations provided by an embodiment of the invention;

fig. 2 is a schematic structural diagram of a ship name and violation identification system according to an embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail below with reference to the accompanying drawings, and it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

In a common system, the ship name recognition is usually carried out only by artificial intelligence, and is influenced by factors such as insufficient illumination at night, indistinguishable multi-ship intersection, large background interference (shop signboards and banners in the background, other characters printed on a ship-borne container), pollution damage or shielding interference (the ship name is shielded by hanging tires and the like), and the recognition rate is low, so that the method is difficult to be truly applied to the illegal recognition. In the traditional method, because the identified ship name content cannot be accurately judged, database inquiry is generally introduced to complement identification, and the requirement of a user on the actual integrity inspection of the ship name content cannot be effectively met.

Thus, the present invention provides a method for identifying a ship name and a violation, and fig. 1 is a flowchart of a method for identifying a ship name and a violation provided in an embodiment of the present invention, as shown in fig. 1, the method includes:

s1, acquiring a panoramic image and a plurality of close-up images of a current ship through a laser identification system, a panoramic camera and a close-up camera; acquiring AIS data of all ships within a specified radius range through an AIS base station, and taking the AIS data as candidate AIS data;

specifically, the S1 includes:

s11, recognizing the position of the ship through a laser recognition system, starting to light at a preset distance before the ship runs to a preset range, and continuously shooting a plurality of close-up images through a close-up camera;

Specifically, the laser recognition system can recognize the relative three-dimensional position of the ship and estimate the speed and direction of the ship. Because the position is not necessarily determined and the night illumination condition is limited, when the position where the ship is about to travel (namely, the ship can appear on the close-up camera picture) can be estimated, the ship starts to light at a plurality of distances in advance, and close-up images of the ship are continuously shot through the close-up camera (the ship is required to be noted, the ship does not need to light in daytime); the time interval for continuous shooting is 1-1.5s to ensure that the shot boat name (i.e. the name on the boat board) is in the middle of the close-up image as much as possible. There may be no ship in the close-up image taken due to the advance by several distances. And ending shooting until the laser identification system displays that the ship leaves the close-up picture for a plurality of distances.

And S12, aligning the central axes of the laser recognition system and the panoramic camera so as to enable the ship to run to the laser prevention area, continuously shooting a plurality of images through the panoramic camera, and taking the image in the middle of shooting time as a panoramic image.

And aligning the central axes of the laser identification system and the panoramic camera, continuously shooting a plurality of images through the panoramic camera when the ship enters the laser defense area, wherein the time interval of continuous shooting is 1-1.5s, ending shooting when the ship leaves the laser defense area, and selecting the image in the middle of shooting time as a panoramic image.

And recording AIS data of all ships within the specified radius range received by the AIS base station as alternative AIS data (such as A data, B data and C data).

Specifically, for example: and when the current ship appears in the laser defense area at 0-49 s, AIS data in a specified radius range at 0-49 s is taken as candidate AIS data.

The AIS data includes: ship name, service identification code, captain, ship type, speed, direction, etc.

S2, judging the ship category according to the panoramic image; performing ship plate recognition on the close-up images, and deleting the invalid images to obtain a plurality of effective ship plate images;

specifically, the type of the ship in the panoramic image is identified as the type of the ship (such as a sand carrier) according to the artificial intelligence; and carrying out ship plate recognition on the multiple close-up images of the current ship through artificial intelligence, and deleting the invalid images to obtain multiple effective ship plate images.

When the ship is specifically identified, the bow, the hull, the stern and the ship plate (marked by the identification frame) can be identified; the ship plate in the bow range is a ship plate in the bow range, the ship plate in the hull range is a ship plate in the hull range, and the ship plate in the stern range is a ship plate in the stern range. When there are multiple bows, multiple hulls, multiple tails and multiple billboards in the close-up image, the bow (containing the billboards), the hull (containing the billboards) and the stern (containing the billboards) of the ship which are not current are filtered, and the bow, the hull and the stern of the ship which are not current are filtered by the relative positions of the ship in the panoramic camera and the close-up camera.

After filtering, at least one of the bow, the hull and the stern of the current ship is required to be contained in the close-up image, otherwise, the close-up image is an invalid image;

the ship plate (the bow ship plate or the hull ship plate or the stern ship plate) is required to be in the close-up image, otherwise, the ship plate is considered as an invalid image;

at least one of four corners of the identification frame of the ship plate is arranged in the bow or the hull or the stern, and the center point of the identification frame of the ship plate is arranged in the bow or the hull or the stern, otherwise, the identification frame of the ship plate is regarded as an invalid image;

center pixel coordinate of identification frame of ship boardDistance from the center pixel coordinates of the close-up image +.>Relative offset +.>If the relative offset is->An invalid image is considered to be greater than a fourth preset threshold;

the maximum length of the ship plate is smaller than a fifth preset threshold value and is regarded as an invalid image; specifically, the maximum length of all the bow and ship plates in the close-up image is obtained, and the bow and ship plates smaller than a fifth preset threshold (for example, 0.8 times of the maximum length of the bow and ship plates) are removed; similarly, the maximum length of all the hull billboards in the close-up image is obtained, and the hull billboards smaller than a fifth preset threshold (for example, 0.8 times of the maximum length of the hull billboards) are removed; and acquiring the maximum length of all the stern billboards in the close-up image, and removing the stern billboards smaller than a fifth preset threshold (for example, 0.8 times of the maximum length of the stern billboards).

And taking the first images with higher ship plate confidence from the plurality of effective ship plate images left after the invalid photos are deleted. Wherein the number of sheets is inversely proportional to the speed of the vessel, and the more rapid the vessel speed, the fewer effective photos the close-up camera can capture.

S3, when the effective ship plate image is judged to be empty, alarming if the ship name is not recognized; otherwise, performing character recognition on the plurality of effective ship board images to obtain a plurality of recognition candidates and confidence degrees corresponding to the recognition candidates; generating a plurality of candidate results according to the identified candidates; performing fuzzy processing and reverse order processing on the candidate result to obtain a plurality of candidate identifications and confidence degrees corresponding to each candidate identification;

and S31, when the effective ship plate image which is not met is found through the judgment of the S2, warning is given by the fact that the ship name is not recognized, and the AIS data which are matched with the current ship in an estimated mode (note that the estimated AIS data which are matched with the current ship are not estimated and are only estimated to be possible AIS information) are provided for a user, wherein the estimated AIS data which are the candidate AIS data with the closest position distance, speed and the like are the same as the current ship direction recognized by the laser recognition system and the direction obtained by the AIS base station. For example: and providing the data A to the user when the direction of the data A in the data A, the data B and the data C is the same as the direction of the current ship identified by the laser identification system and the position in the data A is closest to the position of the current ship identified by the laser identification system.

Otherwise, S321, performing character recognition on a plurality of effective ship board images with higher ship board confidence degrees (which are obtained during artificial intelligence recognition) through artificial intelligence to obtain a plurality of recognition candidates and confidence degrees corresponding to each recognition candidate;

specifically, a plurality of recognition candidates and the confidence corresponding to each recognition candidate (ship name of the ship plate) can be obtained through artificial intelligence recognitionConfidence of individual character in each recognition candidate +.>The recognition frame of each recognition candidate, the recognition frame of a single character in each recognition candidate. Note that, when extracting charactersThe relative position between the characters should not change (i.e., the characters identified in each tile are ordered according to the center pixel coordinates of the identification frame).

For example: identifying candidate Zhejiang shipping 6079) 60791 of Zhejiang boat) "ZheBoyun 6079" ("ZheBoyun")>) 'ZheBo' (-Thunb Bo)>）。

S322, generating a plurality of candidate results according to the identified candidates; performing fuzzy processing and reverse order processing on the candidate result to obtain a plurality of candidate identifications and confidence degrees corresponding to each candidate identification;

the S322 includes:

s3221, combining all words in the recognition candidates to generate a plurality of candidate results;

Specifically, the recognition candidates have five words of "Zhe", "Bo", "voyage", "boat" and "fortune", and according to the relative positions, zhe is known to be in the first position, bo is in the second position, the third position may be voyage/boat/fortune, the fourth position may be voyage/boat/fortune, and the last position is fortune, and the possible candidate results are: "ZheBo shipping", "ZheBo boat shipping".

S3222, adding wildcards into all candidate results to generate fuzzy results so as to carry out fuzzy matching;

specifically, the candidate result is "ZheBo shipping", the corresponding identification candidate is "ZheBo shipping 6079", and the addition of the wild card to the candidate result is "ZheBo% shipping", i.e. "ZheBo% shipping" is likely to be "ZheBo shipping", "ZheBo boat shipping", "ZheBo shipping";

the candidate result is ' ZheBo boat fortune ' and the corresponding identification candidate is ' ZheBo boat fortune 60791 ', and the general match symbol added to the candidate result is ' ZheBo% boat% fortune ', namely ' ZheBo% boat% fortune ' is likely to be ' ZheBo boat fortune ', ' ZheBo;

the candidate result is 'ZheBo fortune' and the corresponding identification candidate is 'ZheBo fortune 6079', and the candidate result is 'ZheBo% fortune' by adding a wildcard, namely 'ZheBo% fortune' is possibly 'ZheBo shipping', 'ZheBo fortune';

The candidate result is ZheBo voyage, the corresponding identification candidate is ZheBo voyage 6079, specifically, zheBo voyage can be derived from ZheBo voyage and ZheBo voyage, and the confidence degree isThe method comprises the steps of carrying out a first treatment on the surface of the The confidence degree of ZheBo boat can be derived from ' navigation ' of ZheBo boat and ' ZheBo boat>The method comprises the steps of carrying out a first treatment on the surface of the Selecting a source route with high confidence, namely a candidate result of ZheBo voyage, zheBo voyage and ZheBo voyage, the corresponding identification candidate is Zhejiang shipping 6079";

the candidate result is ZheBo boat shipping, the corresponding identification candidate is ZheBo boat shipping 6079, specifically, zheBo boat shipping can be derived from ZheBo boat shipping and ZheBo boat shipping, and the confidence degree isThe method comprises the steps of carrying out a first treatment on the surface of the The confidence degree of ZheBo boat can be derived from ' navigation ' of ZheBo boat and ' ZheBo boat>The method comprises the steps of carrying out a first treatment on the surface of the Selecting a source route with high confidence, namely, a ZheBo boat shipping candidate result, namely, zheBo boat shipping with the source of ZheBo boat shipping and a ZheBo boat shipping, wherein the corresponding identification candidates are ZheBo boats Bo-voyage 6079";

in conclusion, the fuzzy results are "ZheBo% voyage", "ZheBo voyage".

S3223, performing reverse order processing on all the fuzzy results to obtain reverse order results;

namely, the reverse sequence result is: "fortune% Bozhe", "fortune boat Bozhe".

S3224, the fuzzy result and the reverse order result are taken as a plurality of candidate identification.

In this example, there are a total of 10 candidate identifications, as shown in the following table:

s4, when all the confidence degrees of the candidate identification are judged not to exceed a first preset threshold value, alarming by the fact that the ship name is not identified; otherwise, sequencing all candidate identifications from high to low according to the confidence level and matching with the candidate AIS data;

specifically, if the confidence of all candidate identifications does not exceed the first preset threshold, warning is given by the unrecognized ship name, and the user is provided with the estimated matched AIS data, (note that the estimated matched AIS data is not estimated to be matched with the current ship, but is only estimated to be possible AIS information), and the estimated matched AIS data is the candidate AIS data with the closest position distance, speed and the like according to the current ship direction identified by the laser identification system and the direction acquired by the AIS base station.

Otherwise, all candidate identifications are ranked from high to low according to the confidence and matched with the candidate AIS data.

S5, when the alternative AIS data are judged to be empty, an AIS alarm is not opened, and a step S6 is carried out; otherwise, calculating the candidate matching degree and the digital matching degree of each candidate identification and the candidate AIS data; when judging that the candidate matching degree does not exceed the second preset threshold value or the digital matching degree does not exceed the third preset threshold value, considering that the ship is not opened with AIS or is covered with cards, and entering step S6; otherwise, the candidate AIS data with the highest candidate matching degree is taken as matching AIS data; judging whether the matched AIS data are matched or not, if so, warning by fake plate; and proceeds to step S7;

specifically, when the candidate AIS data is empty, indicating that the current ship is not opened with AIS, alarming with the non-opened AIS, and entering into step S6;

conversely, in a specific embodiment, calculating candidate matching degrees and digital matching degrees of the 10 candidate identifications and the 3 candidate AIS data respectively;

candidate matching degree =The method comprises the steps of carrying out a first treatment on the surface of the Wherein Sum is the Sum of all words in the current ship board, +.>For the current ship board->Confidence of->For the current ship board->Chinese character->Confidence of the current ship board->Chinese character->When matching is correct +. >=1, matching error ∈>=0；/>Is in all ship plates corresponding to the current ship type c in the (ship name-AIS-ship type) database (the current ship in the panoramic image is identified according to artificial intelligence)Type (2): sand carrier, which is the class of the current ship), word +.>Probability of occurrence at the nth bit, +.>，/>In order to be of the ship type c, the word +.>Number of times appearing at the nth bit; />In order to be of the ship type c, the word +.>The number of occurrences; />Training words in a database for artificial intelligenceFrequency score of (2); namely, in the artificial intelligence training database, the training data words +.>Personal, word->The duty ratio is->；/>Is word->A total number of (3); all characters are based on the duty ratio (>) The sequence of the sequences is carried out,，/>for the ordered median, +.>For the third quartile after ordering, +.>Is the first quartile after sorting. Will->Word +.f. less than a sixth preset threshold (e.g. 0.9 times the first quartile)>=0, will->Word +.f. greater than seventh preset threshold (e.g. third quartile of 1)>=1. The +.>Still calculated->No change is made. Note that the words described above do not include numbers.

Assuming candidate recognition: zheBo% air% operation and alternative AIS data: calculating candidate matching degree by ZheBo boat shipping, wherein the denominator is confidence degree of 5 words of ZheBo, bo, boat and shipping, and the frequency score is obtained when the number of times appears at the nth position in the ship type c; the molecular is confidence of 4 words of Zhe, bo, voyage and fortune, the frequency score of the number of times appearing at the nth position in the ship type c; because of =0。

Digital matching degree =The method comprises the steps of carrying out a first treatment on the surface of the Wherein Sum is the Sum of all numbers in the current ship board, +.>For the current ship board->Confidence of->For the current ship board->Middle number->Confidence of the current ship board->Middle number->When matching is correct +.>=1, matching error ∈>=0；

In particular, when the redundancy of the candidate identified digital portion with respect to the digital portion of the ship name in the candidate AIS data is considered to be likely due to factors such as outboard suspended objects on the ship side or night light reflection, for example, "6079" with respect to "679", the digital match is noted as 100%.

Taking the results (namely the ship name in the alternative AIS data) that the candidate matching degree is highest, the candidate matching degree is higher than a second preset threshold value and the digital matching degree is higher than a third preset threshold value in matching; if the result is empty, the current ship is considered to be not opened with AIS or the sleeved license plate, and the step S6 is carried out;

specifically, after waiting for all ships (refer to the ships corresponding to all the alternative AIS data) to open the specified radius range, if there are remaining unmatched alternative AIS data with the same direction, the current ship is considered to be sleeved; if no unmatched alternative AIS data remain, an AIS alarm is not opened.

The following is a specific example:

If 3 ships (A ship, B ship and C ship) exist in total, the laser identification system identifies the current ship C in the laser defense area, after the ships ABC are all opened out of the specified radius range of the AIS base station, the AIS base station acquires three alternative AIS data (A data, B data and C data) in the specified radius range, and the ship AB is matched with the A data and the B data in the alternative AIS data; if the C data are not matched, but the distance, the speed and the direction of the current ship C are similar to those of the C data in the alternative AIS data, the current ship C is considered to be matched with the C data in the alternative AIS data, and the current ship C is considered to be sleeved;

if there are 4 vessels (A vessel, B vessel, C vessel, D vessel) in total, the laser identification system identifies the current vessel D in the laser defense area, after the vessels ABCD all open the specified radius range of the AIS base station, the AIS base station acquires three alternative AIS data (A data, B data, C data) in the specified radius range, the vessels ABC all match with the A data, the B data and the C data in the alternative AIS data, and no alternative AIS data which is not matched remains, and the current vessel D is considered to not open the AIS.

If the result is not null and the candidate identification is the best match with the C data in the candidate AIS data, modifying the AIS data (such as the A data in the above description) of which the match is estimated initially to be the C data as the matched AIS data; judging whether the matched C data is matched or not, if so, warning by fake plate; for example: the previous ship is matched with the data C in a short time just after the previous ship passes through the data C, and then is matched with the data C after being sleeved by the current ship, and then the ship is warned by the sleeved ship; otherwise, the current ship is considered to be matched with the C data of the alternative AIS data, and the C data is marked as used. Step S7 is entered.

S6, sequencing all candidate identifications from high to low according to the confidence level, matching the candidate identifications with ship names in the database, and calculating the candidate matching degree and the digital matching degree of each candidate identification and the ship name of the database; when the digital matching degree is judged not to exceed a third preset threshold value, warning by a non-registered ship name; otherwise, taking the database ship name with the highest candidate matching degree as a matching ship name, and entering step S7;

sequencing all candidate identifications from high to low according to the confidence level, matching the candidate identifications with ship names in a (ship name-AIS-ship type) database, and calculating the candidate matching degree and the digital matching degree of each candidate identification and the ship name of the database; the calculation method is the same as the above; and (5) taking the result (namely, the name of the ship in the database) that the candidate matching degree is highest and the digital matching degree is higher than a third preset threshold value in matching.

If the result is empty, the default of each word in each candidate recognition is adjusted to be matched with the ship name in the database, and the word in the candidate recognition is preferredIf the name is smaller than the eighth preset threshold value, marking the character as the character, and then matching the character with the ship name in the database; if the result is still empty, e.g. +.of a word in the candidate recognition>=0, but one bit back +.>>0, adding a percent in front of the character, if the result is still empty, alarming by a non-registered ship name, and providing the user with estimated matched AIS data (note that the estimated matched AIS data is not matched with the current ship, but is estimated as possible AIS information), wherein the estimated matched AIS data is the candidate AIS data with the closest position distance, speed and the like according to the current ship direction identified by the laser identification system and the direction acquired by the AIS base station.

Otherwise, taking the database ship name with the highest candidate matching degree as a matching ship name, and entering step S7;

and S7, judging whether the matching degree between the matching AIS data or the matching ship name and the corresponding candidate identification is 100%, if so, obtaining the ship name of the ship, otherwise, shielding by a ship plate and giving an alarm about fouling.

Specifically, matching AIS data or database matching ship names are collectively called as matching results, and corresponding candidate identification and matching results are compared;

for the text part, three situations of word staggering, word missing and multiple words can occur:

1. for candidate recognition few words, check correct word BDeviation distance of->If the value is greater than the threshold value T, the ship plate is considered to possibly have no correct character, and the ship plate is used for shielding and fouling alarm; otherwise, the method is normal;

2. for candidate recognition multiword, check for erroneous word AIs the offset distance of (2)，/>Confidence for A word, ++>For the frequency score of A word, +.>For the probability of the A character appearing in the nth position in the ship type c, if the value is larger than the threshold value, the false character possibly exists on the ship plate, and the ship plate is used for shielding and fouling alarming; otherwise, the method is normal;

3. for candidate recognition error words, calculating the deviation distance of the correct word B and the error word AIf the value is greater than the threshold value, the ship board is considered to be available The false character can be truly existed, and the ship plate is used for shielding and fouling alarm; otherwise, the method is normal.

For the digital part, three cases of word staggering, word leakage and multiple words can occur:

1. for the fewer words of the candidate recognition, checking the interval condition between two numbers in the candidate recognition, if the interval is greater than the average interval of n times, considering that the ship board possibly has no correct word B, and warning by shielding and fouling the ship board, otherwise, the method is normal;

2. for candidate recognition multiple words, errors possibly caused by factors such as outboard suspended objects on the ship board or night light reflection are considered to be normal, such as '6709' relative to '679';

3. for candidate recognition error words, calculating the deviation distance of the correct word B and the error word AIf the value is larger than the threshold value, the false character is considered to exist on the ship board, and the ship board is used for shielding and fouling alarm; otherwise, the method is normal.

Fig. 2 is a schematic structural diagram of a ship name and violation identification system according to an embodiment of the present invention, as shown in fig. 2, the system includes:

an acquisition unit 201 for acquiring a panoramic image and a plurality of close-up images of a current ship through a laser recognition system, a panoramic camera and a close-up camera; acquiring AIS data of all ships within a specified radius range through an AIS base station, and taking the AIS data as candidate AIS data;

The acquisition unit 201 includes:

the first acquisition subunit is used for identifying the position of the ship through the laser identification system, starting to light at a preset distance before the ship runs to a preset range, and continuously shooting a plurality of close-up images through the close-up camera;

and the second acquisition subunit is used for aligning the central axes of the laser identification system and the panoramic camera so as to enable the ship to travel to the laser prevention area, continuously shoot a plurality of images through the panoramic camera, and taking the image in the middle of shooting time as a panoramic image.

A deletion unit 202 for determining a ship category from the panoramic image; performing ship plate recognition on the close-up images, and deleting the invalid images to obtain a plurality of effective ship plate images;

the invalid image includes:

at least one of the bow, the hull and the stern, otherwise, the image is invalid;

the image excluding the ship board is regarded as an invalid image;

at least one of four corners of the identification frame of the ship plate is arranged in the bow or the hull or the stern, and the center point of the identification frame of the ship plate is arranged in the bow or the hull or the stern, otherwise, the identification frame of the ship plate is regarded as an invalid image;

the relative offset of the center pixel coordinate of the identification frame of the ship plate from the center pixel coordinate of the close-up image is larger than a fourth preset threshold value, and the ship plate is regarded as an invalid image;

The maximum length of the ship plate is smaller than the fifth preset threshold value and is regarded as an invalid image.

A first judging unit 203, configured to alarm if no ship name is recognized when it is judged that the effective ship plate image is empty; otherwise, performing character recognition on the plurality of effective ship board images to obtain a plurality of recognition candidates and confidence degrees corresponding to the recognition candidates; generating a plurality of candidate results according to the identified candidates; performing fuzzy processing and reverse order processing on the candidate result to obtain a plurality of candidate identifications and confidence degrees corresponding to each candidate identification;

generating a plurality of candidate results according to the identified candidates; the fuzzy processing and the reverse order processing are carried out on the candidate result, and the obtaining of a plurality of candidate identifications and the confidence degree corresponding to each candidate identification comprises the following steps:

combining all words in the recognition candidates to generate a plurality of candidate results;

adding wildcards into all candidate results to generate fuzzy results so as to carry out fuzzy matching;

performing reverse order processing on all the fuzzy results to obtain reverse order results;

the fuzzy result and the reverse order result are identified as a plurality of candidates.

A second judging unit 204, configured to, when judging that the confidence degrees of all the candidate identifications do not exceed the first preset threshold, alarm with no identification of the ship name; otherwise, sequencing all candidate identifications from high to low according to the confidence level and matching with the candidate AIS data;

A third judging unit 205, configured to, when it is judged that the alternative AIS data is empty, alarm with no AIS open, and enter a database matching unit; otherwise, calculating the candidate matching degree and the digital matching degree of each candidate identification and the candidate AIS data; when the candidate matching degree is judged not to exceed the second preset threshold value or the digital matching degree is judged not to exceed the third preset threshold value, the current ship is considered to be not opened with AIS or is not covered with cards, and the ship is entered into a database matching unit; otherwise, the candidate AIS data with the highest candidate matching degree is taken as matching AIS data; judging whether the matched AIS data are matched or not, if so, warning by fake plate; and then enters a matching degree calculating unit;

when the candidate matching degree is judged not to exceed the second preset threshold value or the digital matching degree is judged not to exceed the third preset threshold value, the current ship is considered to be not opened with AIS or the sleeved license plate comprises:

specifically, after waiting for all ships (refer to the ships corresponding to all the alternative AIS data) to open a specified radius range, if there are residual unmatched alternative AIS data with the same direction, the current ship is considered to be sleeved; if no unmatched alternative AIS data remain, an AIS alarm is not opened.

A database matching unit 206, configured to rank all candidate identifications from high to low according to confidence levels, match the candidate identifications with ship names in the database, and calculate a candidate matching degree and a digital matching degree between each candidate identification and the ship name of the database; when the digital matching degree is judged not to exceed a third preset threshold value, warning by a non-registered ship name; otherwise, taking the database ship name with the highest candidate matching degree as a matching ship name, and entering a matching degree calculation unit;

When the matching degree of the numbers does not exceed the third preset threshold, the warning of the non-registered ship name comprises the following steps:

and when the number matching degree is judged not to exceed a third preset threshold value, adjusting the default of each character in each candidate identification to match with the ship name in the database, and if the matching is still unsuccessful, alarming by using the unregistered ship name.

And the matching degree calculating unit 207 is configured to determine whether the matching degree between the matching AIS data or the matching ship name and the corresponding candidate identification is 100%, if so, obtain the ship name of the ship, otherwise, determine whether the ship board is blocked or stained according to the deviation distance between the matching AIS data or the matching ship name and the corresponding candidate identification.

The invention has the beneficial effects that:

the invention provides a ship name and violation identification method and system, wherein laser is introduced into the method, and fixed-point light filling is carried out at night; processing a multi-vessel scenario by the relative positions of vessels in the panoramic camera and the close-up camera; the artificial intelligence is used for respectively identifying the elements such as the bow, the hull, the stern, the ship plate, the characters and the like so as to avoid the influence of the background as much as possible; introducing alternative AIS data and fuzzy query of a database to defects and shielding, comprehensively analyzing factors such as the occurrence frequency of characters to calculate candidate matching degree and digital matching degree, and judging whether a ship name is sleeved, whether the ship name is not registered and whether the AIS is not opened according to the matching degree score; and finally determining whether the actual candidate recognition ship name accords with the matched result ship name or not through the deviation distance of the candidate recognition ship name and the matched result ship name. The method can accurately identify the ship name and judge the condition of violation.

Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims (10)

1. A method of identifying a ship name and violation, comprising:

s1, acquiring a panoramic image and a plurality of close-up images of a current ship through a laser identification system, a panoramic camera and a close-up camera; acquiring AIS data of all ships within a specified radius range through an AIS base station, and taking the AIS data as candidate AIS data;

s2, judging the ship category according to the panoramic image; performing ship plate recognition on the close-up images, and deleting the invalid images to obtain a plurality of effective ship plate images;

s3, when the effective ship plate image is judged to be empty, alarming if the ship name is not recognized; otherwise, performing character recognition on the plurality of effective ship board images to obtain a plurality of recognition candidates and confidence degrees corresponding to the recognition candidates; generating a plurality of candidate results according to the identified candidates; performing fuzzy processing and reverse order processing on the candidate result to obtain a plurality of candidate identifications and confidence degrees corresponding to each candidate identification;

S4, when all the confidence degrees of the candidate identification are judged not to exceed a first preset threshold value, alarming by the fact that the ship name is not identified; otherwise, sequencing all candidate identifications from high to low according to the confidence level and matching with the candidate AIS data;

s5, when the alternative AIS data are judged to be empty, an AIS alarm is not opened, and a step S6 is carried out; otherwise, calculating the candidate matching degree and the digital matching degree of each candidate identification and the candidate AIS data; when judging that the candidate matching degree does not exceed the second preset threshold value or the digital matching degree does not exceed the third preset threshold value, considering that the current ship is not opened with AIS or is covered with cards, and entering step S6; otherwise, the candidate AIS data with the highest candidate matching degree is taken as matching AIS data; judging whether the matched AIS data are matched or not, if so, warning by fake plate; and proceeds to step S7;

s6, sequencing all candidate identifications from high to low according to the confidence level, matching the candidate identifications with ship names in the database, and calculating the candidate matching degree and the digital matching degree of each candidate identification and the ship name of the database; when the digital matching degree is judged not to exceed a third preset threshold value, warning by a non-registered ship name; otherwise, taking the database ship name with the highest candidate matching degree as a matching ship name, and entering step S7;

S7, judging whether the matching degree between the matching AIS data or the matching ship name and the corresponding candidate identification is 100%, if so, obtaining the ship name of the ship, otherwise, judging whether the ship board is blocked or stained according to the deviation distance between the matching AIS data or the matching ship name and the corresponding candidate identification;

wherein the candidate matching degree is calculated according to the following formula:

candidate matching degree =；

Sum is the Sum of all words in the current ship board,for the current ship board->Confidence of->For the current ship->Chinese character->Confidence of the current ship board->Chinese character->When matching is correct +.>=1, when matching is wrong=0；/>Is the word +.f in all ship plates corresponding to the current ship type c in the database>Probability of occurrence at the nth bit; />，/>In order to be of the ship type c, the word +.>Number of times appearing at the nth bit; />In order to be of the ship type c, the word +.>The number of occurrences; />Training words in a database for artificial intelligenceFrequency score of (2); word +.>Excluding numerals;

the digital matching degree is calculated according to the following formula:

digital matching degree =；

Where Sum is the Sum of all numbers in the current ship board,for the current ship board->Confidence of->For the current ship board->Middle number->Confidence of the current ship board- >Middle number->Matching correct = =>=1, matching error ∈>=0。

2. The method of claim 1, wherein the invalid image comprises:

at least one of the bow, the hull and the stern, otherwise, the image is invalid;

the image excluding the ship board is regarded as an invalid image;

at least one of four corners of the identification frame of the ship plate is arranged in the bow or the hull or the stern, and the center point of the identification frame of the ship plate is arranged in the bow or the hull or the stern, otherwise, the identification frame of the ship plate is regarded as an invalid image;

the relative offset of the center pixel coordinate of the identification frame of the ship plate from the center pixel coordinate of the close-up image is larger than a fourth preset threshold value, and the ship plate is regarded as an invalid image;

the maximum length of the ship plate is smaller than the fifth preset threshold value and is regarded as an invalid image.

3. The method of claim 1, wherein the generating a plurality of candidate results based on identifying candidates; the fuzzy processing and the reverse order processing are carried out on the candidate result, and the obtaining of a plurality of candidate identifications and the confidence degree corresponding to each candidate identification comprises the following steps:

combining all words in the recognition candidates to generate a plurality of candidate results;

adding wildcards into all candidate results to generate fuzzy results so as to carry out fuzzy matching;

Performing reverse order processing on all the fuzzy results to obtain reverse order results;

the fuzzy result and the reverse order result are identified as a plurality of candidates.

4. The method of claim 1, wherein when the candidate matching degree is determined not to exceed the second preset threshold or the digital matching degree is determined not to exceed the third preset threshold, then considering that the current ship is not opened AIS or is being licensed comprises:

after all ships leave the specified radius range, if the rest unmatched alternative AIS data with the same direction exist, the current ship is considered to be sleeved; if no unmatched alternative AIS data remain, an AIS alarm is not opened.

5. The method of claim 1, wherein alerting with a non-registered ship name when the digital match is determined not to exceed a third preset threshold comprises:

and when the number matching degree is judged not to exceed a third preset threshold value, adjusting the default of each character in each candidate identification to match with the ship name in the database, and if the matching is still unsuccessful, alarming by using the unregistered ship name.

6. The method according to claim 1, wherein S1 comprises:

the position of the ship is identified through the laser identification system, and a plurality of close-up images are continuously shot through the close-up camera after the ship starts to light at a preset distance before traveling to a preset range;

And aligning the central axes of the laser identification system and the panoramic camera so as to enable the ship to travel to the laser prevention area, continuously shooting a plurality of images through the panoramic camera, and taking the image in the middle of shooting time as a panoramic image.

7. A ship name and violation identification system, comprising:

the acquisition unit is used for acquiring the panoramic image and a plurality of close-up images of the current ship through the laser identification system, the panoramic camera and the close-up camera; acquiring AIS data of all ships within a specified radius range through an AIS base station, and taking the AIS data as candidate AIS data;

a deleting unit for judging the ship category according to the panoramic image; performing ship plate recognition on the close-up images, and deleting the invalid images to obtain a plurality of effective ship plate images;

the first judging unit is used for alarming when the effective ship plate image is judged to be empty and the ship name is not recognized; otherwise, performing character recognition on the plurality of effective ship board images to obtain a plurality of recognition candidates and confidence degrees corresponding to the recognition candidates; generating a plurality of candidate results according to the identified candidates; performing fuzzy processing and reverse order processing on the candidate result to obtain a plurality of candidate identifications and confidence degrees corresponding to each candidate identification;

The second judging unit is used for alarming when the confidence level of all the candidate identifications is judged not to exceed the first preset threshold value; otherwise, sequencing all candidate identifications from high to low according to the confidence level and matching with the candidate AIS data;

the third judging unit is used for alarming when the alternative AIS data is judged to be empty and entering the database matching unit when the AIS data is not opened; otherwise, calculating the candidate matching degree and the digital matching degree of each candidate identification and the candidate AIS data; when the candidate matching degree is judged not to exceed the second preset threshold value or the digital matching degree is judged not to exceed the third preset threshold value, the current ship is considered to be not opened with AIS or is not covered with cards, and the ship is entered into a database matching unit; otherwise, the candidate AIS data with the highest candidate matching degree is taken as matching AIS data; judging whether the matched AIS data are matched or not, if so, warning by fake plate; and then enters a matching degree calculating unit;

the database matching unit is used for sequencing all candidate identifications from high to low according to the confidence level, matching the candidate identifications with the ship names in the database, and calculating the candidate matching degree and the digital matching degree of each candidate identification and the ship name of the database; when the digital matching degree is judged not to exceed a third preset threshold value, warning by a non-registered ship name; otherwise, taking the database ship name with the highest candidate matching degree as a matching ship name, and entering a matching degree calculation unit;

The matching degree calculation unit is used for judging whether the matching degree between the matching AIS data or the matching ship name and the corresponding candidate identification is 100%, if so, the ship name of the ship is obtained, otherwise, whether the ship board is blocked or stained is judged according to the deviation distance between the matching AIS data or the matching ship name and the corresponding candidate identification;

wherein the candidate matching degree is calculated according to the following formula:

candidate matching degree =；

Sum is the Sum of all words in the current ship board,for the current ship board->Confidence of->Is the current ship boardChinese character->Confidence of the current ship board->Chinese character->When matching is correct +.>=1, when matching is wrong=0；/>Is the word +.f in all ship plates corresponding to the current ship type c in the database>Probability of occurrence at the nth bit; />，/>In order to be of the ship type c, the word +.>Number of times appearing at the nth bit; />In order to be of the ship type c, the word +.>The number of occurrences; />Training words in a database for artificial intelligenceFrequency score of (2); word +.>Excluding numerals;

the digital matching degree is calculated according to the following formula:

digital matching degree =；

Where Sum is the Sum of all numbers in the current ship board,for the current ship board->Confidence of->For the current ship board->Middle number- >Confidence of the current ship board->Middle number->When matching is correct +.>=1, matching error ∈>=0。

8. The system of claim 7, wherein the generating a plurality of candidate results based on identifying candidates; the fuzzy processing and the reverse order processing are carried out on the candidate result, and the obtaining of a plurality of candidate identifications and the confidence degree corresponding to each candidate identification comprises the following steps:

combining all words in the recognition candidates to generate a plurality of candidate results;

adding wildcards into all candidate results to generate fuzzy results so as to carry out fuzzy matching;

performing reverse order processing on all the fuzzy results to obtain reverse order results;

the fuzzy result and the reverse order result are identified as a plurality of candidates.

9. The system of claim 7, wherein when the candidate matching degree is determined not to exceed the second preset threshold or the digital matching degree is determined not to exceed the third preset threshold, then considering that the current ship is not opened AIS or is being licensed comprises:

after all ships leave the specified radius range, if the rest unmatched alternative AIS data with the same direction exist, the current ship is considered to be sleeved; if no unmatched alternative AIS data remain, an AIS alarm is not opened.

10. The system of claim 7, wherein alerting with a non-registered ship name when the digital match is determined not to exceed a third preset threshold comprises:

And when the number matching degree is judged not to exceed a third preset threshold value, adjusting the default of each character in each candidate identification to match with the ship name in the database, and if the matching is still unsuccessful, alarming by using the unregistered ship name.