CN111144475A - Method and device for determining car seat, electronic equipment and readable storage medium - Google Patents
Method and device for determining car seat, electronic equipment and readable storage medium Download PDFInfo
- Publication number
- CN111144475A CN111144475A CN201911354665.1A CN201911354665A CN111144475A CN 111144475 A CN111144475 A CN 111144475A CN 201911354665 A CN201911354665 A CN 201911354665A CN 111144475 A CN111144475 A CN 111144475A
- Authority
- CN
- China
- Prior art keywords
- seat
- confidence
- seats
- predicted
- positions
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 52
- 238000012216 screening Methods 0.000 claims abstract description 71
- 238000001514 detection method Methods 0.000 claims abstract description 41
- 238000013135 deep learning Methods 0.000 claims abstract description 26
- 230000015654 memory Effects 0.000 claims description 17
- 238000012549 training Methods 0.000 claims description 11
- 238000004590 computer program Methods 0.000 claims description 7
- 238000012163 sequencing technique Methods 0.000 claims description 6
- 238000001914 filtration Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 abstract description 11
- 230000007547 defect Effects 0.000 abstract 1
- 238000012986 modification Methods 0.000 description 10
- 230000004048 modification Effects 0.000 description 10
- 238000012550 audit Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 6
- 238000003491 array Methods 0.000 description 5
- 239000003550 marker Substances 0.000 description 5
- 238000007619 statistical method Methods 0.000 description 5
- 238000007689 inspection Methods 0.000 description 4
- 230000005291 magnetic effect Effects 0.000 description 4
- 230000006870 function Effects 0.000 description 3
- 230000000717 retained effect Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- 230000003936 working memory Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/40—Business processes related to the transportation industry
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Computational Biology (AREA)
- Economics (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Image Analysis (AREA)
Abstract
The application discloses a method and a device for determining a carriage seat, electronic equipment and a readable storage medium. The method for determining the carriage seat comprises the following steps: acquiring a carriage image to be detected; adopting a seat target detection model based on deep learning to obtain the seat category, confidence and predicted position of each seat in the carriage image; screening the predicted positions of all the seats according to the seat types and the confidence degrees to obtain target seats; and determining the carriage seat according to the category and the predicted position of the target seat. The application also discloses a device for determining the carriage seat, electronic equipment and a readable storage medium; the method can automatically and accurately determine the carriage seat, and can avoid the defects of high labor cost, low efficiency and low accuracy in the process of determining the carriage seat.
Description
Technical Field
The present disclosure relates to the field of vehicle detection, and in particular, to a method and an apparatus for determining a car seat, an electronic device, and a readable storage medium.
Background
With the rapid development of social economy and the increasing improvement of living standard of people, the holding quantity of urban motor vehicles is rapidly increased at present, which directly brings about the remarkable increase of the vehicle inspection workload in the vehicle transaction process.
The inventors found that at least the following problems exist in the related art: in the conventional vehicle inspection, the verification work of the carriage seat (such as the carriage seat of a passenger car) mainly depends on manual operation, the carriage seat needs to be determined manually, and whether the corresponding vehicle is modified or not is checked based on the determined carriage seat. In the process of determining the carriage seat, the labor cost is high, the checking efficiency is low, and long-time repeated checking operation easily causes manual entering into a fatigue state, so that the checking accuracy is easily influenced due to negligence. Therefore, how to automatically and accurately determine the carriage seat and avoid the disadvantages of high labor cost, low efficiency and low accuracy in the process of determining the carriage seat are technical problems which need to be solved in the current situation.
Disclosure of Invention
An object of the present application is to provide a method, an apparatus, an electronic device and a readable storage medium for determining a car seat, which can automatically and accurately determine the car seat, and can avoid the disadvantages of high labor cost, low efficiency and low accuracy in the process of determining the car seat.
According to one aspect of the present application, there is provided a method of determining a car seat, including: acquiring a carriage image to be detected; adopting a seat target detection model based on deep learning to obtain the seat category, confidence and predicted position of each seat in the carriage image; screening the predicted positions of all the seats according to the seat types and the confidence degrees to obtain target seats; and determining the carriage seat according to the category and the predicted position of the target seat.
According to another aspect of the present application, there is also provided a car seat detecting device, including: the first acquisition module is used for acquiring a carriage image to be detected; the second acquisition module is used for acquiring the seat category, the confidence coefficient and the predicted position of each seat in the carriage image by adopting a seat target detection model based on deep learning; the screening module is used for screening the predicted positions of all the seats according to the seat types and the confidence degrees to obtain target seats; and the determining module is used for determining the carriage seat according to the category and the predicted position of the target seat.
According to another aspect of the present application, there is also provided an electronic device including: one or more processors; and a memory storing computer readable instructions that, when executed, cause the processor to perform the above-described car seat determination method.
According to another aspect of the present application, there is also provided a computer-readable storage medium storing a computer program which, when executed by a processor, implements the above-described method for determining a vehicle seat.
In addition, the screening the predicted position of each seat according to the seat category and the confidence to obtain a target seat includes: obtaining the confidence of each seat in the same seat type in each seat type; sequencing the seats according to the sequence of the confidence degrees from high to low; and screening the predicted positions of the seats according to the arrangement sequence of the seats to obtain the target seat.
In addition, the screening the predicted positions of the seats according to the arrangement sequence of the seats to obtain the target seat includes: obtaining the residual prediction positions except the prediction position with the highest confidence coefficient; according to the sequence of the confidence degrees of the residual prediction positions from high to low, the overlapping degree of the prediction position with the highest confidence degree and the residual prediction positions is calculated one by one; traversing all the residual prediction positions according to a preset screening strategy; according to the sequence of the confidence degrees of the predicted positions obtained after screening from high to low, the overlapping degree of the predicted position with the second highest confidence degree and the predicted position of each seat obtained after screening is calculated one by one; traversing all the residual prediction positions obtained after the screening according to the preset screening strategy; repeating the steps until the overlapping degree of the prediction position with the second last confidence ranking and the prediction position with the lowest confidence ranking is calculated according to the sequence of the confidence degrees of the residual prediction positions from high to low, and screening according to the preset screening strategy to obtain the final residual prediction position; and determining a target seat according to the final residual predicted position.
In addition, traversing all the remaining predicted positions according to a preset screening strategy comprises: when the calculated overlapping degree is larger than a preset threshold value, deleting the predicted position with low confidence coefficient in the two currently compared predicted positions, and continuing to execute traversal operation until all the rest seats are traversed; and when the calculated overlapping degree is smaller than or equal to the preset threshold, simultaneously keeping the corresponding predicted positions of the two seats which are currently compared, and continuously executing traversal operation until all the rest seats are traversed.
In addition, the deleting, when the calculated overlap degree is greater than a preset threshold, a predicted position with a low confidence coefficient from among two predicted positions currently being compared includes: when the calculated overlapping degree is larger than a preset threshold value, calculating the score of the prediction position with low confidence coefficient in the two current prediction positions to be compared according to the overlapping degree; and if the score is smaller than a preset score, executing the step of deleting the predicted position with low confidence coefficient in the two currently compared predicted positions, otherwise, respectively keeping the two currently compared predicted positions.
In addition, the deep learning-based seat target detection model is obtained by the following specific steps: adjusting the VGG basic model; training to obtain the seat target detection model based on deep learning according to the adjusted VGG basic model; wherein the adjusting the VGG base model comprises: acquiring a mean file of each seat; according to the mean file, the output result of the SSD framework is modified, and the basic learning rate is set to 0.0001, the weight attenuation value is set to 0.0005, the learning rate policy is set to multistep, the display parameter is set to 0.1, and the momentum is set to 0.9.
In addition, after the determining the carriage seat according to the category and the predicted position of the target seat, the method further comprises the following steps: judging whether the carriage seat is consistent with a pre-stored carriage seat or not, and if so, judging that the carriage seat is not modified; otherwise, judging that the refitting of the carriage seat exists.
The application has at least the following beneficial effects:
1. the predicted positions of the seats are screened according to the seat types and the confidence degrees, and the carriage seats are determined according to the types and the predicted positions of the target seats obtained after screening, so that the carriage seats can be automatically and accurately determined, subsequent auditing work is facilitated, the labor cost can be saved, and the accuracy rate of determining the carriage seats and the efficiency of determining the carriage seats are improved.
2. The method can be applied to the determination of the seats of the passenger car in the annual inspection of the vehicle, so that the overall auditing time is shortened, the current requirements on the working efficiency and accuracy of the annual inspection of the vehicle can be met, and the waiting time of the owner of the motor vehicle can be reduced.
Description of the drawings:
one or more embodiments are illustrated by way of example in the accompanying drawings, which correspond to the figures in which like reference numerals refer to similar elements and which are not to scale unless otherwise specified.
FIG. 1 is a flow chart of a method of determining a car seat provided in accordance with an aspect of the present application;
fig. 2 is a schematic diagram illustrating screening of predicted positions of individual seats in a method for determining a car seat according to an aspect of the present application;
fig. 3 is a schematic diagram illustrating a method for determining a car seat according to an aspect of the present application, wherein the predicted positions of two seats are filtered;
fig. 4 is a schematic diagram illustrating a method for determining a car seat according to an aspect of the present application, wherein the predicted positions of three seats are filtered.
The specific implementation mode is as follows:
in order to make the objects, technical solutions and advantages of the embodiments of the present invention more apparent, embodiments of the present invention will be described in detail below based on the accompanying drawings. However, it will be appreciated by those of ordinary skill in the art that numerous technical details are set forth in order to provide a better understanding of the present application in various embodiments of the present invention. However, the technical solution claimed in the present application can be implemented without these technical details and various changes and modifications based on the following embodiments.
Fig. 1 shows a flow chart of a method for determining a car seat according to an aspect of the present application, the method comprising steps 101 to 104:
in step 101, a compartment image to be detected is acquired; here, the image of the vehicle compartment to be detected may be acquired from the imaging device. In some embodiments, the car image must include the plus driver seat, the secondary driver seat, and all other seats.
In step 102, a seat target detection model based on deep learning is adopted to obtain the seat category, confidence and predicted position of each seat in the carriage image; after a seat target detection model based on deep learning receives a carriage image to be detected, N one-dimensional arrays [ class, x, y, width, height ] can be obtained, wherein N is a natural number which is greater than or equal to 1; the first element class in the array represents the class of the seat, for example, if the class is a single seat, the class value is 1, if the class is a double seat, the class value is 2, if the class is a triple seat, the class value is 3, and if the class is not a seat, the class value is 0; the last four elements in the array represent a rectangular area where the seat is located, namely the predicted position in the embodiment, x and y respectively represent the abscissa and the ordinate of the upper left corner point of the rectangular area, width represents the width of the rectangular area, and height represents the height of the rectangular area; then, based on the values of the elements in the one-dimensional array, a confidence score corresponding to each one-dimensional array is obtained, and the confidence score is used to represent the probability of the possibility that a rectangular region is the actual position of a seat. It should be noted that each array corresponds to a predicted position (e.g., a single seat, a double seat, a triple seat). In this embodiment, whether a seat exists in the car image may be detected by using a seat target detection model based on deep learning, and if so, the detected seat is extracted, and the seat type, the confidence level, and the predicted position of the detected seat are obtained, and if not, the process is ended.
In some embodiments, the deep learning based seat target detection model may be specifically obtained by the following means, including:
preparing training data: acquiring interior images of a carriage, including a front driving seat, a secondary driving seat and all other seats, shot by different vehicle types, different brands and appointed shooting angles; herein, the vehicle type may include, but is not limited to: 5 seat surface charters, 7 seat surface charters, 9 passenger cars, 11 seat surface charters and business cars;
labeling data: marking the position of each seat in the image in the carriage and the seat type to which each seat belongs by adopting a rectangular frame; here, the seats included within the area marked by the rectangular box should be complete seats.
Training a model: and training to obtain a seat target detection model by using the labeled images. In some embodiments, a VGG (Visual Geometry Group) base model trained by ImageNet may be used, the labeled image may be input into a VGG framework, and fine-tuning may be performed on the VGG base model to obtain a seat target detection model through training.
It should be noted that the seat categories in this embodiment may include, but are not limited to: at least one of the single seat, the double seat and the three seats.
When the seat category includes only a single seat, the deep learning-based seat target detection model may be a single seat target detection model, which may be specifically obtained as follows: acquiring interior images of a carriage, including a front driving seat, a secondary driving seat and all other seats, shot by different vehicle types, different brands and appointed shooting angles; adopting a rectangular frame to mark the position of the single seat; and training to obtain the single-seat target detection model by using the marked images.
When the seat category includes only two seats, the deep learning-based seat target detection model may be a two-seat target detection model, which may be specifically obtained as follows: acquiring interior images of a carriage, including a front driving seat, a secondary driving seat and all other seats, shot by different vehicle types, different brands and appointed shooting angles; adopting a rectangular frame to mark the positions of the two seats; and training to obtain the double-seat target detection model by using the marked images.
When the seat category includes only three seats, the deep learning-based seat target detection model may be a three-seat target detection model, which may be specifically obtained as follows: acquiring interior images of a carriage, including a front driving seat, a secondary driving seat and all other seats, shot by different vehicle types, different brands and appointed shooting angles; adopting a rectangular frame to mark the positions of the three seats; and training to obtain the three-seat target detection model by using the marked images.
In step 103, screening the predicted positions of the seats according to the seat categories and the confidence degrees to obtain target seats; here, reference may be made to fig. 2, 3, and 4, where fig. 2 is a schematic diagram illustrating screening of predicted positions of a single seat, fig. 3 is a schematic diagram illustrating screening of predicted positions of two seats, and fig. 4 is a schematic diagram illustrating screening of predicted positions of three seats. Examples of the seat categories include: single-seat and double-seat, then, the predicted position of each single-seat can be screened according to the confidence of each single-seat, and the predicted position of each double-seat can be screened according to the confidence of each double-seat. It can be understood that, in the output result of the seat target detection model based on deep learning, a plurality of predicted positions may be obtained from the same seat, for example, although predicted position 1 and predicted position 2 belong to two different predicted positions, the predicted positions 1 and predicted position 2 may both represent the same seat a, and therefore, the predicted positions of the seats need to be screened to obtain the target seat.
In step 104, the car seat is determined based on the category and predicted position of the target seat. Here, after the predicted positions of the seats are screened, the predicted positions of the single seats, the predicted positions of the double seats, and/or the predicted positions of the triple seats can be determined, and the number and distribution of the seats in each seat type in the vehicle compartment seat can be determined. In some embodiments, the distribution of seats inside the carriage can be constructed according to the area size of the rectangular box for representing the predicted position and the relative position of each seat under different seat categories, and the seats of the carriage are determined.
In an embodiment of the present application, in step 103, the screening the predicted position of each seat according to the seat category and the confidence level to obtain a target seat includes: obtaining the confidence of each seat in the same seat type in each seat type; sequencing the seats according to the sequence of the confidence degrees from high to low; and screening the predicted positions of the seats according to the arrangement sequence of the seats to obtain the target seat. In some embodiments, the predicted position corresponding to the confidence coefficient smaller than the preset confidence coefficient in the confidence coefficients may be deleted. For example, if there are 5 predicted positions detected in a single seat category, the confidence levels are: 0.9, 0.5, 0.8, 0.3 and 0.6, and obtaining the following results after sequencing the seats according to the sequence of confidence degrees from high to low: 0.9, 0.8, 0.6, 0.5 and 0.3, if the preset confidence coefficient is 0.5, the predicted position with the confidence coefficient of 0.3 can be deleted, and the target seat can be obtained. In some embodiments, the sorted predicted positions of the seats may be further filtered with reference to the standard number of the single seats, and assuming that the standard number of the single seats is 4, in the above example, the predicted positions corresponding to 0.9, 0.8, 0.6, and 0.5 in the sorted seats may be retained, and the predicted position with the lowest confidence level, that is, the predicted position with the confidence level of 0.3, may be deleted to obtain the target seat. Assuming that the standard number of the single seat is 3, in the above example, the predicted positions corresponding to 0.9, 0.8, and 0.6 in each seat after sorting may be retained, and the last two bits of the confidence ranking, that is, the predicted positions with the confidence of 0.3 and 0.5, may be deleted to obtain the target seat.
Continuing with the foregoing embodiment, the screening the predicted positions of the seats according to the arrangement order of the seats to obtain the target seat may include: obtaining the residual prediction positions except the prediction position with the highest confidence coefficient; according to the sequence of the confidence degrees of the residual prediction positions from high to low, the overlapping degree of the prediction position with the highest confidence degree and the residual prediction positions is calculated one by one; traversing all the residual prediction positions according to a preset screening strategy;
according to the sequence of the confidence degrees of the predicted positions obtained after screening from high to low, the overlapping degree of the predicted position with the second highest confidence degree and the predicted position of each seat obtained after screening is calculated one by one; traversing all the residual prediction positions obtained after the screening according to the preset screening strategy; repeating the steps until the overlapping degree of the prediction position with the second last confidence ranking and the prediction position with the lowest confidence ranking is calculated according to the sequence of the confidence degrees of the residual prediction positions from high to low, and screening according to the preset screening strategy to obtain the final residual prediction position; and determining a target seat according to the final residual predicted position.
Here, for example, if the number of detected predicted positions in a single seat category is 5, the confidence levels are: 0.9, 0.5, 0.8, 0.3 and 0.6, and obtaining the following results after sequencing the seats according to the sequence of confidence degrees from high to low: 0.9, 0.8, 0.6, 0.5, and 0.3, the remaining predicted positions except the predicted position with the highest confidence coefficient, that is, the remaining predicted positions except the predicted position with the confidence coefficient of 0.9, may be obtained first, and the confidence coefficients of the remaining predicted positions are: 0.8, 0.6, 0.5, 0.3. Then, according to the order of the confidence degrees of the residual prediction positions from high to low, the overlapping degree of the prediction position with the highest confidence degree and the residual prediction positions is calculated one by one, namely the overlapping degree of the prediction position with the confidence degree of 0.9 and the prediction position with the confidence degree of 0.8, the overlapping degree of the prediction position with the confidence degree of 0.9 and the prediction position with the confidence degree of 0.6, the overlapping degree of the prediction position with the confidence degree of 0.9 and the prediction position with the confidence degree of 0.5, and the overlapping degree of the prediction position with the confidence degree of 0.9 and the prediction position with the confidence degree of 0.3, so that all the residual prediction positions can be calculated to be traversed. Assuming that all the remaining predicted positions are traversed according to a preset screening strategy, the screened predicted positions include: a predicted position with a confidence of 0.9, a predicted position with a confidence of 0.8, a predicted position with a confidence of 0.5, a predicted position with a confidence of 0.3.
Then, according to the sequence of the confidence degrees of the predicted positions obtained after screening from high to low, the predicted positions with the second highest confidence degrees, namely the overlapping degree of the predicted position with the confidence degree of 0.8 and the predicted position of each seat obtained after screening, are calculated one by one, namely: the degree of overlap between the predicted position with the confidence of 0.8 and the predicted position with the confidence of 0.9, the degree of overlap between the predicted position with the confidence of 0.8 and the predicted position with the confidence of 0.5, and the degree of overlap between the predicted position with the confidence of 0.8 and the predicted position with the confidence of 0.3. It is worth mentioning that in practical applications, the degree of overlap between the predicted position with the confidence level of 0.8 and the predicted position with the confidence level of 0.9 has already been calculated in the previous step, and may not be calculated any more thereafter. Assuming that, according to a preset screening strategy, after traversing all the remaining predicted positions obtained after the screening, the predicted positions obtained after the screening include: a predicted position with a confidence of 0.9, a predicted position with a confidence of 0.8, a predicted position with a confidence of 0.5.
By analogy, in this example, the confidence levels of the remaining predicted positions are: 0.9, 0.8 and 0.5. In this case, the predicted position with the confidence of 0.8 is the predicted position with the second lowest confidence rank, and the predicted position with the confidence of 0.5 is the predicted position with the first lowest confidence rank, so that the final residual predicted position can be obtained after the degree of overlap between the predicted position with the confidence of 0.8 and the predicted position with the confidence of 0.5 is obtained. That is, the final remaining predicted positions are: a predicted position with a confidence of 0.9, a predicted position with a confidence of 0.8, a predicted position with a confidence of 0.5.
In some embodiments, the confidence levels detected in the single seat categories in the above examples may be: the 5 predicted positions of 0.9, 0.5, 0.8, 0.3, 0.6 are stored in the first vector V1 in order of confidence from high to low, denoted as [0.9, 0.8, 0.6, 0.5, 0.3], and the final remaining predicted positions are stored in the first vector V2 in order of confidence from high to low, denoted as [0.9, 0.8, 0.5 ].
In some embodiments, after receiving a car image to be detected, a deep learning-based seat target detection model may obtain N one-dimensional arrays [ x, y, width, height ] for a seat with a seat category specifically being a single seat, where four elements in the arrays represent a rectangular region where the single seat is located, that is, a predicted position in this embodiment, x and y represent an abscissa and an ordinate of an upper left corner point of the rectangular region, respectively, width represents a width of the rectangular region, and height represents a height of the rectangular region; determining the area of the rectangular area according to the value of each element in the one-dimensional array, so that the rectangular area with the area larger than the preset area can be subjected to non-maximum value suppression treatment, and the confidence coefficient corresponding to the rectangular area with the largest set area is highest; and screening the rectangular areas (namely the predicted positions) of the single seats according to the preset screening strategy. Similarly, when the seat type is specifically a double seat and/or a triple seat, the processing method of the deeply learned seat target detection model after receiving the car image to be detected is substantially the same as the processing method when the seat type is specifically a single seat, and in order to avoid repetition, the description is not repeated here, and only the differences are explained below:
optionally, in the case that the seat category is specifically a single seat, since the single seat is generally a front driver seat, a front passenger seat, and a two-row small single seat, the single seat may be further screened according to the relative positions of each single seat and the seats of other seat categories, so as to delete the rectangular area representing the single seat that is unlikely to appear in the rear row of the vehicle compartment.
Optionally, for the case that the seat category is specifically two seats and/or three seats, if the values of the ordinate y of the upper left corner point of the rectangular areas in the plurality of one-dimensional arrays are approximately the same, it may be determined that the plurality of rectangular areas corresponding to the plurality of one-dimensional arrays are in the same horizontal state, and a merging operation may be performed on the plurality of rectangular areas, so that the rectangular areas are merged into a more complete rectangular frame.
Optionally, for the case that the seat category is specifically double seats and/or triple seats, since the characteristic difference between the double seats and the triple seats is often not obvious, the number of the double seats and the triple seats can be adjusted by calculating the difference between the value of the width of the rectangular region of the double seats and the value of the width of the rectangular region of the triple seats, and determining whether the width is the width of a single seat, so that the accuracy of the detection result can be further improved.
Continuing with the above embodiment, the traversing all the remaining predicted positions according to the preset filtering policy may include: when the calculated overlapping degree is larger than a preset threshold value, deleting the predicted position with low confidence coefficient in the two currently compared predicted positions, and continuing to execute traversal operation until all the rest seats are traversed; and when the calculated overlapping degree is smaller than or equal to the preset threshold, simultaneously keeping the corresponding predicted positions of the two seats which are currently compared, and continuously executing traversal operation until all the rest seats are traversed. Wherein, the overlapping degree may be specifically: the ratio of the area of the intersection region to the area of the phase-parallel region in the two prediction positions currently being compared.
Here, for example, if the number of detected predicted positions in a single seat category is 3, the seats are sorted according to the order of confidence level from high to low, and then: 0.9, 0.8 and 0.6. First, the degree of overlap between the predicted position with confidence of 0.9 and the predicted position with confidence of 0.8, and the overlap between the predicted position with confidence of 0.9 and the predicted position with confidence of 0.6 are compared, respectively. When the degree of overlap between the predicted position with the confidence coefficient of 0.9 and the predicted position with the confidence coefficient of 0.8 is greater than the preset threshold, it can be stated to some extent that the predicted position with the confidence coefficient of 0.9 and the predicted position with the confidence coefficient of 0.8 have a larger overlap region, and the two predicted positions are likely to indicate the same seat, so that the predicted position with high confidence coefficient, i.e., the predicted position with the confidence coefficient of 0.9, is retained, and the predicted position with low confidence coefficient, i.e., the predicted position with the confidence coefficient of 0.8, is deleted. When the degree of overlap between the predicted position with the confidence coefficient of 0.9 and the predicted position with the confidence coefficient of 0.8 is less than or equal to the preset threshold, it can be stated that the predicted position with the confidence coefficient of 0.9 and the predicted position with the confidence coefficient of 0.8 have no overlapping region or have a small overlapping region, and the two predicted positions are likely to indicate two independent seats, so that the predicted position with the confidence coefficient of 0.9 and the predicted position with the confidence coefficient of 0.8 which are currently compared are simultaneously reserved. Similarly, the predicted position with the confidence coefficient of 0.9 and the predicted position with the confidence coefficient of 0.6, and the predicted position with the confidence coefficient of 0.8 and the predicted position with the confidence coefficient of 0.6 are screened by the same screening strategy.
It can be understood that the method for determining the seats of the carriage in the embodiment can effectively solve the problem of misidentification that three seats include two seats and two seats include a single seat under the condition that seats of various types of the carriage are relatively similar, and improve the accuracy of the detection result.
Continuing with the foregoing embodiment, the deleting, when the calculated overlap degree is greater than the preset threshold, the predicted position with low confidence coefficient from the two predicted positions currently being compared may include: when the calculated overlapping degree is larger than a preset threshold value, calculating the score of the prediction position with low confidence coefficient in the two current prediction positions to be compared according to the overlapping degree; and if the score is smaller than a preset score, executing the step of deleting the predicted position with low confidence coefficient in the two currently compared predicted positions, otherwise, respectively keeping the two currently compared predicted positions. The specific value of the preset score is set according to the actual situation, and may be 0.1, for example.
Here, calculating the score of the predicted position with low confidence in the two currently compared predicted positions according to the overlap degree may include: and calculating the score of the predicted position with low confidence coefficient according to the overlapping degree and the confidence coefficient of the predicted position with low confidence coefficient. Score (1-overlap) of predicted positions with low confidence, for example; where score represents the confidence of the predicted location with low confidence, overlap represents the overlap of the two predicted locations currently being compared. In this embodiment, by performing a further strict screening on the deletion operation of the predicted position with low confidence in the two currently compared predicted positions, it can be solved that, in some cases in the prior art, although the overlap degree is greater than the preset threshold, the predicted position with low confidence in the two currently compared predicted positions may also be used for representing a seat independent from the predicted position with low confidence, so as to further avoid erroneous judgment, and further improve the accuracy of the detection result.
In an embodiment of the present application, in step 102, the deep learning-based seat target detection model is obtained by: adjusting the VGG basic model; training to obtain the seat target detection model based on deep learning according to the adjusted VGG basic model; wherein the adjusting the VGG base model comprises:
acquiring a mean file of each seat; according to the mean file, the output result of the SSD framework is modified, and the basic learning rate is set to 0.0001, the weight attenuation value is set to 0.0005, the learning rate policy is set to multistep, the display parameter is set to 0.1, and the momentum is set to 0.9.
In an embodiment of the present application, after step 104, that is, after determining the car seat according to the category and the predicted position of the target seat, the method may further include: judging whether the carriage seat is consistent with a pre-stored carriage seat or not, and if so, judging that the carriage seat is not modified; otherwise, judging that the refitting of the carriage seat exists.
In an actual scene application of the present application, a method for determining a car seat includes:
s1, downloading the image of the compartment to be detected and the archived seat information from the server; the seat information referred to herein may include, but is not limited to, seat category, actual location, number of seats under each seat category.
S2, judging whether a seat exists in the carriage image by adopting a seat target detection model based on deep learning, if so, recording the flag bit as first flag information, such as 0, extracting the detected seat, and entering the next step; if the current compartment image does not exist, recording the flag bit as second flag information, such as 1, storing the current compartment image, and entering a statistical analysis process;
s3, detecting the seat type, confidence and predicted position of the seat by adopting a seat target detection model based on deep learning;
s4, judging whether a main driver seat and a secondary driver seat exist for the seats with the single seat type, if so, screening according to the relative positions of the single seats and the seats of other seat types respectively to delete the rectangular areas of the single seats which are not possibly arranged at the rear row of the carriage and are used for representing the single seats, and screening the predicted positions according to the confidence degrees of the single seats to obtain a target seat; if not, the predicted positions can be directly screened according to the confidence of each single seat to obtain the target seat. After the target seats are obtained, calculating the number of the target seats of the single seat, comparing the number with the number of the single seat archived by the server, if the number is consistent with the number of the single seat, recording that the mark position is first mark information, such as 0, extracting the image of the single seat, and entering the next step; if the two images are not consistent, recording the flag bit as second flag information, such as 1, and storing the current carriage image, wherein the current carriage image is a problem image, and entering a statistical analysis process;
and S5, for the seats with the seat types of double seats, screening the predicted positions according to the confidence degrees of the double seats to obtain the target seats. After the target seats are obtained, calculating the number of the target seats of the double seats, comparing the number with the number of the double seats archived by the server, if the number is consistent with the number of the double seats archived by the server, recording that the mark position is first mark information, such as 0, extracting the images of the double seats, and entering the next step; if the two images are not consistent, recording the flag bit as second flag information, such as 1, and storing the current carriage image, wherein the current carriage image is a problem image, and entering a statistical analysis process;
and S6, for the seats with the seat types of three, screening the predicted positions according to the confidence degrees of the three seats to obtain the target seats. After the target seats are obtained, calculating the number of the target seats of the three seats, comparing the number with the number of the three seats archived by the server, if the number is consistent with the number of the three seats, recording that the mark position is first mark information, such as 0, extracting the images of the three seats, and entering the next step; if the two images are not consistent, recording the flag bit as second flag information, such as 1, and storing the current carriage image, wherein the current carriage image is a problem image, and entering a statistical analysis process;
s7, carrying out statistical analysis on the marker information of the whole process, and if all the recorded marker bits are first marker information, such as 0, judging that the carriage seat is not modified and passes the examination; and if the recorded marker bit has second marker information, such as 1, judging that the coach seat is refitted, and not passing the audit. In addition, the specific link is determined to be qualified according to the position where the first mark information appears, and the specific link is determined to be unqualified according to the position where the second mark information appears.
The method for determining the carriage seat can automatically audit the carriage seat, so that the existing manual audit mode can be replaced, the labor cost is saved, the audit speed is increased, and the justness and the disclosure of the audit work are ensured.
In another practical application scenario of the present application, the car seat may be determined as follows: a one-dimensional array [ x1, x2, x3, x4, x5] may be used to indicate the audit state, with the initial values set to [0, 0, 0, 0, 0], where,
the flag x1 represents whether a seat exists in the car image, and if so, the value of x1 is 0, and if not, the value of x1 is 1;
the flag x2 represents whether the detected number of single seats is consistent with the number of single seats archived by the server, if so, the value of x2 is 0, and if not, the value of x2 is 1;
the flag x3 represents whether the detected number of double seats is consistent with the number of double seats archived by the server, if so, the value of x3 is 0, and if not, the value of x3 is 1;
the flag x4 represents whether the detected number of three seats is consistent with the number of three seats archived by the server, if so, the value of x4 is 0, and if not, the value of x4 is 1;
the flag x5 represents whether the predicted position of the seat of each seat type matches the seat distribution of each seat type stored in the server, and if so, the value of x5 is 0, and if not, the value of x5 is 1.
Finally, counting the states of the zone bits, if the values of the zone bits are all 0, judging that the carriage seat is not modified, and checking to pass; and if the value of at least one flag bit is 1, judging that the carriage seat is refitted, and not passing the audit.
Optionally, the reason why the audit fails may be automatically analyzed further according to the position where the flag bit value is 1:
if the value of x1 is 1, it indicates that there is no seat in the car image, there is a refit in the car seat, or that the car image is not captured at a proper angle, such as: the car image does not include the entire seat, or the car image is taken from a rear position of the car.
If the value of x2 is 1, it indicates that there is a refit in the car seat, or the car image is not captured at a specified angle, such as: the front driver seat and the front passenger seat in the car image are not shot, so that the images can be sent to relevant workers for the relevant workers to check.
If the value of x3 is 1, it indicates that there is a modification of the car seat, such as a possibility of modifying the double seat into another type of seat, or removing the double seat.
If the value of x4 is 1, it indicates that there is a modification of the car seat, such as the possibility of modifying three seats into other types of seats, or removing three seats.
If the value of x5 is 1, then a modification of the car seat is present, for example, a modification of the car seat may be performed separately.
If x5 is 1, the car seat distribution is modified.
In addition, in an embodiment of the present application, there is also provided a car seat detection apparatus including: the first acquisition module is used for acquiring a carriage image to be detected; the second acquisition module is used for acquiring the seat category, the confidence coefficient and the predicted position of each seat in the carriage image by adopting a seat target detection model based on deep learning; the screening module is used for screening the predicted positions of all the seats according to the seat types and the confidence degrees to obtain target seats; and the determining module is used for determining the carriage seat according to the category and the predicted position of the target seat.
In some embodiments, said screening the predicted position of each seat according to the seat category and the confidence level to obtain a target seat includes: obtaining the confidence of each seat in the same seat type in each seat type; sequencing the seats according to the sequence of the confidence degrees from high to low; and screening the predicted positions of the seats according to the arrangement sequence of the seats to obtain the target seat.
In some embodiments, the screening the predicted positions of the seats according to the arrangement order of the seats to obtain the target seat includes: obtaining the residual prediction positions except the prediction position with the highest confidence coefficient; according to the sequence of the confidence degrees of the residual prediction positions from high to low, the overlapping degree of the prediction position with the highest confidence degree and the residual prediction positions is calculated one by one; traversing all the residual prediction positions according to a preset screening strategy; according to the sequence of the confidence degrees of the predicted positions obtained after screening from high to low, the overlapping degree of the predicted position with the second highest confidence degree and the predicted position of each seat obtained after screening is calculated one by one; traversing all the residual prediction positions obtained after the screening according to the preset screening strategy; repeating the steps until the overlapping degree of the prediction position with the second last confidence ranking and the prediction position with the lowest confidence ranking is calculated according to the sequence of the confidence degrees of the residual prediction positions from high to low, and screening according to the preset screening strategy to obtain the final residual prediction position; and determining a target seat according to the final residual predicted position.
In some embodiments, traversing all the remaining predicted positions according to the preset filtering policy includes: when the calculated overlapping degree is larger than a preset threshold value, deleting the predicted position with low confidence coefficient in the two currently compared predicted positions, and continuing to execute traversal operation until all the rest seats are traversed; and when the calculated overlapping degree is smaller than or equal to the preset threshold, simultaneously keeping the corresponding predicted positions of the two seats which are currently compared, and continuously executing traversal operation until all the rest seats are traversed.
In some embodiments, the deleting the predicted position with low confidence from the two currently compared predicted positions when the calculated overlap is greater than the preset threshold includes: when the calculated overlapping degree is larger than a preset threshold value, calculating the score of the prediction position with low confidence coefficient in the two current prediction positions to be compared according to the overlapping degree; and if the score is smaller than a preset score, executing the step of deleting the predicted position with low confidence coefficient in the two currently compared predicted positions, otherwise, respectively keeping the two currently compared predicted positions.
In some embodiments, the deep learning based seat target detection model is obtained by: adjusting the VGG basic model; training to obtain the seat target detection model based on deep learning according to the adjusted VGG basic model; wherein the adjusting the VGG base model comprises: acquiring a mean file of each seat; according to the mean file, the output result of the SSD framework is modified, and the basic learning rate is set to 0.0001, the weight attenuation value is set to 0.0005, the learning rate policy is set to multistep, the display parameter is set to 0.1, and the momentum is set to 0.9.
In some embodiments, after said determining the car seat based on the category and the predicted position of the target seat, further comprises: judging whether the carriage seat is consistent with a pre-stored carriage seat or not, and if so, judging that the carriage seat is not modified; otherwise, judging that the refitting of the carriage seat exists.
It should be understood that the present embodiment is an apparatus embodiment corresponding to the above-described embodiment of the method for determining a vehicle seat, and the present embodiment can be implemented in cooperation with the embodiment of the method for determining a vehicle seat. The related technical details mentioned in the embodiments of the car seat determination method are still valid in the embodiments of the present application, and are not described herein again in order to reduce repetition.
In an embodiment of the present application, there is also provided an electronic device, including: one or more processors; and a memory storing computer readable instructions that, when executed, cause the processor to perform a method of determining a car seat as in any one of the above.
The embodiment of the application also provides a computer readable medium, and the computer program is used for realizing any one of the above-mentioned car seat determination methods when being executed by a processor.
For example, the computer readable instructions, when executed, cause the one or more processors to:
acquiring a carriage image to be detected;
adopting a seat target detection model based on deep learning to obtain the seat category, confidence and predicted position of each seat in the carriage image;
screening the predicted positions of all the seats according to the seat types and the confidence degrees to obtain target seats;
and determining the carriage seat according to the category and the predicted position of the target seat.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present application without departing from the spirit and scope of the application. Thus, if such modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is intended to include such modifications and variations as well.
It should be noted that the present application may be implemented in software and/or a combination of software and hardware, for example, implemented using Application Specific Integrated Circuits (ASICs), general purpose computers or any other similar hardware devices. In one embodiment, the software programs of the present application may be executed by a processor to implement the steps or functions described above. Likewise, the software programs (including associated data structures) of the present application may be stored in a computer readable recording medium, such as RAM memory, magnetic or optical drive or diskette and the like. Additionally, some of the steps or functions of the present application may be implemented in hardware, for example, as circuitry that cooperates with the processor to perform various steps or functions.
In addition, some of the present application may be implemented as a computer program product, such as computer program instructions, which when executed by a computer, may invoke or provide methods and/or techniques in accordance with the present application through the operation of the computer. Program instructions which invoke the methods of the present application may be stored on a fixed or removable recording medium and/or transmitted via a data stream on a broadcast or other signal-bearing medium and/or stored within a working memory of a computer device operating in accordance with the program instructions. An embodiment according to the present application comprises an apparatus comprising a memory for storing computer program instructions and a processor for executing the program instructions, wherein the computer program instructions, when executed by the processor, trigger the apparatus to perform a method and/or a solution according to the aforementioned embodiments of the present application.
It will be evident to those skilled in the art that the present application is not limited to the details of the foregoing illustrative embodiments, and that the present application may be embodied in other specific forms without departing from the spirit or essential attributes thereof. The present embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the application being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference sign in a claim should not be construed as limiting the claim concerned. Furthermore, it is obvious that the word "comprising" does not exclude other elements or steps, and the singular does not exclude the plural. The terms first, second, etc. are used to denote names, but not any particular order.
In a typical configuration of the present application, the terminal, the device serving the network, and the trusted party each include one or more processors (e.g., Central Processing Units (CPUs)), input/output interfaces, network interfaces, and memory.
The Memory may include volatile Memory in a computer readable medium, Random Access Memory (RAM), and/or nonvolatile Memory such as Read Only Memory (ROM) or flash Memory (flash RAM). Memory is an example of a computer-readable medium.
Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to, Phase-Change RAM (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), flash Memory or other Memory technology, Compact Disc Read-Only Memory (CD-ROM), Digital Versatile Disc (DVD) or other optical storage, magnetic cassette tape, magnetic tape storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include non-transitory computer readable media (transmyedia), such as modulated data signals and carrier waves.
The basic principles and the main features of the solution and the advantages of the solution have been shown and described above. It will be understood by those skilled in the art that the present solution is not limited to the embodiments described above, which are described in the specification and illustrated only to illustrate the principles of the solution, but that various changes and modifications may be made to the solution without departing from the spirit and scope of the solution, and these changes and modifications are intended to be within the scope of the claimed solution. The scope of the present solution is defined by the appended claims and equivalents thereof.
Claims (10)
1. A method for determining a vehicle seat, comprising:
acquiring a carriage image to be detected;
adopting a seat target detection model based on deep learning to obtain the seat category, confidence and predicted position of each seat in the carriage image;
screening the predicted positions of all the seats according to the seat types and the confidence degrees to obtain target seats;
and determining the carriage seat according to the category and the predicted position of the target seat.
2. The method for determining a car seat according to claim 1, wherein the step of screening the predicted position of each seat according to the seat category and the confidence level to obtain a target seat comprises:
obtaining the confidence of each seat in the same seat type in each seat type;
sequencing the seats according to the sequence of the confidence degrees from high to low;
and screening the predicted positions of the seats according to the arrangement sequence of the seats to obtain the target seat.
3. A method for determining a car seat according to claim 2, wherein the step of screening the predicted position of each seat according to the arrangement sequence of each seat to obtain a target seat comprises:
obtaining the residual prediction positions except the prediction position with the highest confidence coefficient;
according to the sequence of the confidence degrees of the residual prediction positions from high to low, the overlapping degree of the prediction position with the highest confidence degree and the residual prediction positions is calculated one by one;
traversing all the residual prediction positions according to a preset screening strategy;
according to the sequence of the confidence degrees of the predicted positions obtained after screening from high to low, the overlapping degree of the predicted position with the second highest confidence degree and the predicted position of each seat obtained after screening is calculated one by one;
traversing all the residual prediction positions obtained after the screening according to the preset screening strategy;
repeating the steps until the overlapping degree of the prediction position with the second last confidence ranking and the prediction position with the lowest confidence ranking is calculated according to the sequence of the confidence degrees of the residual prediction positions from high to low, and screening according to the preset screening strategy to obtain the final residual prediction position;
and determining a target seat according to the final residual predicted position.
4. A method for determining a car seat as claimed in claim 3, wherein said traversing all the remaining predicted positions according to the predetermined filtering strategy comprises:
when the calculated overlapping degree is larger than a preset threshold value, deleting the predicted position with low confidence coefficient in the two currently compared predicted positions, and continuing to execute traversal operation until all the rest seats are traversed;
and when the calculated overlapping degree is smaller than or equal to the preset threshold, simultaneously keeping the corresponding predicted positions of the two seats which are currently compared, and continuously executing traversal operation until all the rest seats are traversed.
5. The method for determining a car seat according to claim 4, wherein the deleting the predicted position with low confidence in the two predicted positions currently being compared when the calculated degree of overlap is greater than a preset threshold value comprises:
when the calculated overlapping degree is larger than a preset threshold value, calculating the score of the prediction position with low confidence coefficient in the two current prediction positions to be compared according to the overlapping degree;
and if the score is smaller than a preset score, executing the step of deleting the predicted position with low confidence coefficient in the two currently compared predicted positions, otherwise, respectively keeping the two currently compared predicted positions.
6. The method for determining a car seat according to claim 1, wherein the deep learning-based seat target detection model is obtained by:
adjusting the VGG basic model;
training to obtain the seat target detection model based on deep learning according to the adjusted VGG basic model; wherein the adjusting the VGG base model comprises:
acquiring a mean file of each seat; according to the mean file, the output result of the SSD framework is modified, and the basic learning rate is set to 0.0001, the weight attenuation value is set to 0.0005, the learning rate policy is set to multistep, the display parameter is set to 0.1, and the momentum is set to 0.9.
7. A car seat determination method as claimed in any one of claims 1 to 6, further comprising, after said determining the car seat based on the category and predicted position of the target seat:
judging whether the carriage seat is consistent with the pre-stored carriage seat,
if the seats are consistent, judging that the seats of the carriage are not modified; otherwise, judging that the refitting of the carriage seat exists.
8. A carriage seat detection apparatus, comprising:
the first acquisition module is used for acquiring a carriage image to be detected;
the second acquisition module is used for acquiring the seat category, the confidence coefficient and the predicted position of each seat in the carriage image by adopting a seat target detection model based on deep learning;
the screening module is used for screening the predicted positions of all the seats according to the seat types and the confidence degrees to obtain target seats;
and the determining module is used for determining the carriage seat according to the category and the predicted position of the target seat.
9. An electronic device, comprising:
one or more processors; and
memory storing computer readable instructions which, when executed, cause the processor to perform a method of determining a car seat as claimed in any one of claims 1 to 7.
10. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out a method for determining a car seat according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911354665.1A CN111144475A (en) | 2019-12-22 | 2019-12-22 | Method and device for determining car seat, electronic equipment and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911354665.1A CN111144475A (en) | 2019-12-22 | 2019-12-22 | Method and device for determining car seat, electronic equipment and readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111144475A true CN111144475A (en) | 2020-05-12 |
Family
ID=70519861
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911354665.1A Pending CN111144475A (en) | 2019-12-22 | 2019-12-22 | Method and device for determining car seat, electronic equipment and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111144475A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112000808A (en) * | 2020-09-29 | 2020-11-27 | 迪爱斯信息技术股份有限公司 | Data processing method and device and readable storage medium |
Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108009544A (en) * | 2017-12-13 | 2018-05-08 | 北京小米移动软件有限公司 | Object detection method and device |
| CN109145756A (en) * | 2018-07-24 | 2019-01-04 | 湖南万为智能机器人技术有限公司 | Object detection method based on machine vision and deep learning |
| CN109145854A (en) * | 2018-08-31 | 2019-01-04 | 东南大学 | A kind of method for detecting human face based on concatenated convolutional neural network structure |
| CN109344835A (en) * | 2018-09-20 | 2019-02-15 | 上海眼控科技股份有限公司 | Altering detecting method based on vehicle VIN code character position |
| CN109343701A (en) * | 2018-09-03 | 2019-02-15 | 电子科技大学 | A kind of intelligent human-machine interaction method based on dynamic hand gesture recognition |
| CN109711262A (en) * | 2018-11-28 | 2019-05-03 | 大连理工大学 | A kind of intelligent excavator pedestrian detection method based on depth convolutional neural networks |
| CN109902676A (en) * | 2019-01-12 | 2019-06-18 | 浙江工业大学 | A kind of separated based on dynamic background stops detection algorithm |
| CN109902576A (en) * | 2019-01-25 | 2019-06-18 | 华中科技大学 | A kind of training method and application of head shoulder images classifier |
| CN109934121A (en) * | 2019-02-21 | 2019-06-25 | 江苏大学 | A kind of orchard pedestrian detection method based on YOLOv3 algorithm |
| CN110111331A (en) * | 2019-05-20 | 2019-08-09 | 中南大学 | Honeycomb paper core defect inspection method based on machine vision |
| CN110223302A (en) * | 2019-05-08 | 2019-09-10 | 华中科技大学 | A kind of naval vessel multi-target detection method extracted based on rotary area |
| CN110245691A (en) * | 2019-05-27 | 2019-09-17 | 上海眼控科技股份有限公司 | A kind of intelligent identification Method of vehicle appearance color discoloration repacking |
| CN110378258A (en) * | 2019-07-04 | 2019-10-25 | 上海眼控科技股份有限公司 | A kind of vehicle seat information detecting method and equipment based on image |
| CN110530779A (en) * | 2019-08-09 | 2019-12-03 | 上海依智医疗技术有限公司 | A kind of white blood cell detection method and device |
-
2019
- 2019-12-22 CN CN201911354665.1A patent/CN111144475A/en active Pending
Patent Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108009544A (en) * | 2017-12-13 | 2018-05-08 | 北京小米移动软件有限公司 | Object detection method and device |
| CN109145756A (en) * | 2018-07-24 | 2019-01-04 | 湖南万为智能机器人技术有限公司 | Object detection method based on machine vision and deep learning |
| CN109145854A (en) * | 2018-08-31 | 2019-01-04 | 东南大学 | A kind of method for detecting human face based on concatenated convolutional neural network structure |
| CN109343701A (en) * | 2018-09-03 | 2019-02-15 | 电子科技大学 | A kind of intelligent human-machine interaction method based on dynamic hand gesture recognition |
| CN109344835A (en) * | 2018-09-20 | 2019-02-15 | 上海眼控科技股份有限公司 | Altering detecting method based on vehicle VIN code character position |
| CN109711262A (en) * | 2018-11-28 | 2019-05-03 | 大连理工大学 | A kind of intelligent excavator pedestrian detection method based on depth convolutional neural networks |
| CN109902676A (en) * | 2019-01-12 | 2019-06-18 | 浙江工业大学 | A kind of separated based on dynamic background stops detection algorithm |
| CN109902576A (en) * | 2019-01-25 | 2019-06-18 | 华中科技大学 | A kind of training method and application of head shoulder images classifier |
| CN109934121A (en) * | 2019-02-21 | 2019-06-25 | 江苏大学 | A kind of orchard pedestrian detection method based on YOLOv3 algorithm |
| CN110223302A (en) * | 2019-05-08 | 2019-09-10 | 华中科技大学 | A kind of naval vessel multi-target detection method extracted based on rotary area |
| CN110111331A (en) * | 2019-05-20 | 2019-08-09 | 中南大学 | Honeycomb paper core defect inspection method based on machine vision |
| CN110245691A (en) * | 2019-05-27 | 2019-09-17 | 上海眼控科技股份有限公司 | A kind of intelligent identification Method of vehicle appearance color discoloration repacking |
| CN110378258A (en) * | 2019-07-04 | 2019-10-25 | 上海眼控科技股份有限公司 | A kind of vehicle seat information detecting method and equipment based on image |
| CN110530779A (en) * | 2019-08-09 | 2019-12-03 | 上海依智医疗技术有限公司 | A kind of white blood cell detection method and device |
Non-Patent Citations (2)
| Title |
|---|
| 廖佳伟: "基于卷积神经网络的无人机目标检测算法研究", vol. 2019, no. 2, pages 031 - 525 * |
| 陈慧岩: "《智能车辆理论与应用》", pages: 71 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112000808A (en) * | 2020-09-29 | 2020-11-27 | 迪爱斯信息技术股份有限公司 | Data processing method and device and readable storage medium |
| CN112000808B (en) * | 2020-09-29 | 2024-04-16 | 迪爱斯信息技术股份有限公司 | Data processing method and device and readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110378258B (en) | Image-based vehicle seat information detection method and device | |
| CN109086668B (en) | Unmanned aerial vehicle remote sensing image road information extraction method based on multi-scale generation countermeasure network | |
| CN110348360B (en) | Detection report identification method and equipment | |
| JP6757758B2 (en) | How to split images of inspection equipment and vehicles | |
| CN110414421B (en) | Behavior identification method based on continuous frame images | |
| CN110765963A (en) | Vehicle brake detection method, device, equipment and computer readable storage medium | |
| CN108805864A (en) | The acquisition methods and device of architecture against regulations object based on view data | |
| CN110991437B (en) | Character recognition method and device, training method and device for character recognition model | |
| CN110796078A (en) | Vehicle light detection method and device, electronic equipment and readable storage medium | |
| CN112613387A (en) | Traffic sign detection method based on YOLOv3 | |
| CN117173568A (en) | Target detection model training method and target detection method | |
| CN110807493A (en) | Optimization method and equipment of vehicle classification model | |
| CN116964588A (en) | Target detection method, target detection model training method and device | |
| CN111881984A (en) | Target detection method and device based on deep learning | |
| CN109523557B (en) | Image semantic segmentation labeling method, device and storage medium | |
| CN111144475A (en) | Method and device for determining car seat, electronic equipment and readable storage medium | |
| CN114463300A (en) | Steel surface defect detection method, electronic device, and storage medium | |
| CN111709951B (en) | Target detection network training method and system, network, device and medium | |
| CN113221853A (en) | Yolov 4-based chicken farm feeding identification algorithm | |
| CN111126286A (en) | Vehicle dynamic detection method and device, computer equipment and storage medium | |
| CN116486129A (en) | Deep learning-based railway wagon cover plate fault identification method and device | |
| CN113033720B (en) | Vehicle bottom picture foreign matter identification method and device based on sliding window and storage medium | |
| CN115526855A (en) | Method and device for detecting subfissure defect of battery piece, computer equipment and storage medium | |
| CN111127701B (en) | Vehicle failure scene detection method and system | |
| CN113326891A (en) | Method, system and device for detecting small target object |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20200512 |
|
| WD01 | Invention patent application deemed withdrawn after publication |