WO2020259291A1 - 指示灯的指示信息识别方法及装置、电子设备和存储介质 - Google Patents

指示灯的指示信息识别方法及装置、电子设备和存储介质 Download PDF

Info

Publication number
WO2020259291A1
WO2020259291A1 PCT/CN2020/095437 CN2020095437W WO2020259291A1 WO 2020259291 A1 WO2020259291 A1 WO 2020259291A1 CN 2020095437 W CN2020095437 W CN 2020095437W WO 2020259291 A1 WO2020259291 A1 WO 2020259291A1
Authority
WO
WIPO (PCT)
Prior art keywords
target object
indicator
area
classifier
detection result
Prior art date
Application number
PCT/CN2020/095437
Other languages
English (en)
French (fr)
Inventor
马佳彬
何哲琪
王坤
曾星宇
Original Assignee
商汤集团有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 商汤集团有限公司 filed Critical 商汤集团有限公司
Priority to SG11202102205TA priority Critical patent/SG11202102205TA/en
Priority to JP2021512798A priority patent/JP2022500739A/ja
Priority to KR1020217009669A priority patent/KR20210052525A/ko
Publication of WO2020259291A1 publication Critical patent/WO2020259291A1/zh
Priority to US17/194,175 priority patent/US20210192239A1/en

Links

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
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/09—Arrangements for giving variable traffic instructions
    • G08G1/0962—Arrangements for giving variable traffic instructions having an indicator mounted inside the vehicle, e.g. giving voice messages
    • 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
    • G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
    • G06F18/24147—Distances to closest patterns, e.g. nearest neighbour classification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/285—Selection of pattern recognition techniques, e.g. of classifiers in a multi-classifier system
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
    • G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
    • G06V20/584—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads of vehicle lights or traffic lights
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/09—Arrangements for giving variable traffic instructions
    • G08G1/0962—Arrangements for giving variable traffic instructions having an indicator mounted inside the vehicle, e.g. giving voice messages
    • G08G1/09623—Systems involving the acquisition of information from passive traffic signs by means mounted on the vehicle
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/09—Arrangements for giving variable traffic instructions
    • G08G1/0962—Arrangements for giving variable traffic instructions having an indicator mounted inside the vehicle, e.g. giving voice messages
    • G08G1/0967—Systems involving transmission of highway information, e.g. weather, speed limits
    • G08G1/096708—Systems involving transmission of highway information, e.g. weather, speed limits where the received information might be used to generate an automatic action on the vehicle control
    • G08G1/096725—Systems involving transmission of highway information, e.g. weather, speed limits where the received information might be used to generate an automatic action on the vehicle control where the received information generates an automatic action on the vehicle control
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/07—Target detection

Definitions

  • the present disclosure relates to the field of computer vision technology, and in particular to a method and device for identifying indicator information of indicator lights, electronic equipment and storage media.
  • Traffic lights are devices that are set on roads to provide guidance signals for vehicles and pedestrians. Road conditions are very complicated, and sudden changes or accidents may occur at any time. Traffic lights can adjust many contradictions and prevent accidents by adjusting the passage time of different objects. For example, at an intersection, vehicles in different lanes will preempt to pass through the intersection, causing conflicts.
  • traffic lights can be used in different scenes, have different shapes and types, and have complex undertaking relationships.
  • the present disclosure proposes a technical solution for identifying indicator information of indicator lights.
  • a method for identifying indication information of an indicator light which includes:
  • a detection result of a target object is determined, the target object includes at least one of an indicator light base and an indicator light in a lighting state, and the detection result includes the type of the target object and the input image The position of the target area where the target object is located;
  • the target area in the input image where the target object is located is recognized to obtain the indication information of the target object.
  • the determining the detection result of the target object based on the input image includes:
  • the intermediate detection result of each candidate area including the prediction type of the target object and the target object being The prediction probability of the prediction type;
  • the prediction type is any one of the indicator lamp base and the N types of indicator lamps in the lighting state, and N is a positive integer;
  • the detection result of the target object is determined based on the intermediate detection result of each candidate region in the at least one candidate region and the first position of each candidate region.
  • determining the intermediate detection result of each candidate area based on the image feature at the first position corresponding to each candidate area in the input image includes:
  • the preset type includes at least one of an indicator lamp base and N types of indicator lamps with lighting states, and N is a positive integer;
  • the preset type with the highest prediction probability among the at least one preset type is used as the prediction type of the target object in the candidate area, and the prediction probability of the prediction type is obtained.
  • the method before determining the detection result of the target object based on the intermediate detection result of each candidate region in the at least one candidate region and the first position of each candidate region, the method further includes:
  • the determining the detection result of the target object based on the intermediate detection result of each candidate region in the at least one candidate region and the first position of each candidate region includes:
  • the predicted type of the target object in the target area is taken as the type of the target object
  • the first position of the target area is taken as the position of the target area where the target object is located
  • the detection result of the target object is obtained .
  • the method further includes at least one of the following:
  • the state of the scene in which the input image is collected is the dark state.
  • the recognizing the target region where the target object is located in the input image based on the detection result of the target object to obtain the indication information of the target object includes:
  • the matching classifier is used to identify the image features of the target area in the input image to obtain the indication information of the target object.
  • the recognizing the target region where the target object is located in the input image based on the detection result of the target object to obtain the indication information of the target object includes:
  • the classifier that is determined to match includes a first classifier for identifying the arrangement of the indicator lights in the indicator light base, using the first classifier, Identify the image features of the target area where the target object is located, and determine the arrangement of the indicator lights in the indicator light base; and/or,
  • the classifier that determines the match includes a second classifier for recognizing the scene where the indicator is located, and using the second classifier to recognize the image feature of the target area where the target object is located, and determine the indicator Information about the scene where the light is located.
  • the recognizing the target region where the target object is located in the input image based on the detection result of the target object to obtain the indication information of the target object includes:
  • the matching classifier includes a third classifier for identifying the color attribute of the round-spot light
  • the third classifier is used to identify the image feature of the target area where the target object is located, and determine the color attribute of the round spot light or the pedestrian light.
  • the recognizing the target region where the target object is located in the input image based on the detection result of the target object to obtain the indication information of the target object includes:
  • the classifier determined to match includes a fourth classifier for identifying the color attribute of the arrow light and a fifth classifier for identifying the direction attribute;
  • the image features of the target area where the target object is located are recognized, and the color attribute and the direction attribute of the arrow light are respectively determined.
  • the recognizing the target region where the target object is located in the input image based on the detection result of the target object to obtain the indication information of the target object includes:
  • the classifier determined to match includes a sixth classifier for identifying the color attribute of the digital lamp and a seventh classifier for identifying the numerical attribute;
  • the image features of the target area where the target object is located are recognized, and the color attribute and the numerical attribute of the digital lamp are determined respectively.
  • the method in response to the input image including at least two indicator light bases, the method further includes:
  • the first indicator light base determine an indicator light matching the first indicator light base in a lighting state; the first indicator light base is one of the at least two indicator light bases;
  • the determining the indicator light matching the lighting state of the first indicator light base includes:
  • the ratio between the second area and the second area of the first indicator lamp in the lit state is greater than the setting An area threshold, which determines that the first indicator light in the lighting state matches the first indicator light base;
  • the first indicator lamp in the lighting state is one of the at least one indicator lamp in the lighting state.
  • a driving control method which includes:
  • a control instruction of the intelligent driving device is generated.
  • an indicating information identification device of an indicator light which includes:
  • An acquisition module which is used to acquire an input image
  • a detection module which is used to determine a detection result of a target object based on the input image, the target object includes at least one of an indicator lamp base and an indicator lamp in a lighting state, and the detection result includes the target object's Type, the position of the target area where the target object is located in the input image;
  • the recognition module is used for recognizing the target area where the target object is located in the input image based on the detection result of the target object to obtain the indication information of the target object.
  • the determining module is further used for:
  • the intermediate detection result of each candidate area including the prediction type of the target object and the target object being The prediction probability of the prediction type;
  • the prediction type is any one of the indicator lamp base and the N types of indicator lamps in the lighting state, and N is a positive integer;
  • the detection result of the target object is determined based on the intermediate detection result of each candidate region in the at least one candidate region and the first position of each candidate region.
  • the determining module is further configured to: for each candidate area, classify the target object in the candidate area based on the image feature at the first position corresponding to the candidate area to obtain The target object is the predicted probability of each preset type in the at least one preset type; wherein, the preset type includes at least one of an indicator light base and N types of indicator lights in the lighting state, N Is a positive integer;
  • the preset type with the highest prediction probability among the at least one preset type is used as the prediction type of the target object in the candidate area, and the prediction probability of the prediction type is obtained.
  • the determining module is further configured to: before determining the detection result of the target object based on the intermediate detection result of each candidate region in the at least one candidate region and the first position of each candidate region , Determining the position deviation of the first position of each candidate region based on the image feature of the input image;
  • the determining module is further configured to, when there are at least two candidate regions of the target object, based on the intermediate detection result of each of the at least two candidate regions, or based on The intermediate detection result of each candidate area and the first position of each candidate area, selecting the target area from the at least two candidate areas;
  • the predicted type of the target object in the target area is taken as the type of the target object
  • the first position of the target area is taken as the position of the target area where the target object is located
  • the detection result of the target object is obtained .
  • the determining module is further configured to determine that the indicator light is in a fault state when the detection result of the target object only includes the detection result corresponding to the indicator light base;
  • the detection result of the target object only includes the detection result corresponding to the indicator lamp in the lit state, it is determined that the state of the scene in which the input image is collected is the dark state.
  • the recognition module is further configured to determine a classifier matching the target object based on the type of the target object in the detection result of the target object;
  • the matching classifier is used to identify the image features of the target area in the input image to obtain the indication information of the target object.
  • the identification module is further configured to determine that the matching classifier includes an arrangement for arranging the indicator lights in the indicator light base when the type of the target object is an indicator light base
  • the first classifier that recognizes in a way, using the first classifier to recognize the image features of the target area where the target object is located, and determine the arrangement of the indicator lights in the indicator light base;
  • the classifier that determines the match includes a second classifier for recognizing the scene where the indicator is located, and using the second classifier to recognize the image feature of the target area where the target object is located, and determine the indicator Information about the scene where the light is located.
  • the recognition module is further configured to determine that the matching classifier includes a classifier used to identify the color attribute of the round spot light when the type of the target object is a round spot light or a pedestrian light.
  • the third classifier is further configured to determine that the matching classifier includes a classifier used to identify the color attribute of the round spot light when the type of the target object is a round spot light or a pedestrian light.
  • the third classifier is used to identify the image feature of the target area where the target object is located, and determine the color attribute of the round spot light or the pedestrian light.
  • the recognition module is further configured to determine that the matching classifier includes a fourth classifier and direction for determining the color attribute of the arrow lamp when the type of the target object is an arrow light.
  • the image features of the target area where the target object is located are recognized, and the color attribute and the direction attribute of the arrow light are respectively determined.
  • the recognition module is further configured to determine that the matching classifier includes a sixth classifier and a numerical value for the color attribute of the digital lamp when the type of the target object is a digital lamp.
  • the image features of the target area where the target object is located are recognized, and the color attribute and the numerical attribute of the digital lamp are determined respectively.
  • the device further includes a matching module, which is used to determine a match with the first indicator base for the first indicator base when the input image includes at least two indicator bases.
  • the matching module is also used to:
  • the ratio of the first area between the first indicator lamp in the lit state and the first indicator lamp base to the second area of the first indicator lamp in the lit state is greater than the set area In the case of a threshold, it is determined that the first indicator light in the lighting state matches the first indicator light base;
  • the first indicator lamp in the lighting state is one of the at least one indicator lamp in the lighting state.
  • a driving control device which includes:
  • An image acquisition module which is set in the smart driving device and used to collect driving images of the smart driving device
  • An image processing module configured to execute the indicator light indication information recognition method as described in any one of the first aspect on the driving image to obtain the indication information for the driving image;
  • the control module is configured to use the instruction information to generate a control instruction of the intelligent driving device.
  • an electronic device including:
  • a memory for storing processor executable instructions
  • the processor is configured to call an instruction stored in the memory to execute the method of any one of the first aspect or the second aspect.
  • a computer-readable storage medium having computer program instructions stored thereon, and when the computer program instructions are executed by a processor, the computer program instructions in the first aspect or the second aspect are implemented.
  • a computer program including computer readable code, and when the computer readable code is executed in an electronic device, a processor in the electronic device executes for realizing the first aspect Or the method of any one of the second aspect.
  • the embodiments of the present disclosure may first perform target detection processing on the input image to obtain the target object detection result, where the target object detection result may include information such as the position and type of the target object, and then further execute the target object instruction according to the target object detection result Identification of information.
  • the present disclosure divides the detection process of the target object into two detection processes of the indicator light base and the indicator light in the lighting state, and realizes the first distinction of the target object during the detection process, and then performs subsequent detection based on the detection result of the target object.
  • Fig. 1 shows a flowchart of a method for identifying indicator information of an indicator light according to an embodiment of the present disclosure
  • Figure 2(a) shows different display states of traffic lights
  • FIG. 2(b) shows different arrangements of traffic light bases
  • Figure 2(c) shows different application scenarios of traffic lights
  • Figure 2(d) shows various types of traffic lights
  • Figure 2(e) shows a schematic diagram of combined traffic lights in different situations
  • FIG. 3 shows a flowchart of step S20 in a method for identifying indicator information of an indicator light according to an embodiment of the present disclosure
  • FIG. 4 shows a schematic diagram of performing target detection through a regional candidate network according to an embodiment of the present disclosure
  • FIG. 5 shows a flowchart of step S30 in a method for identifying indication information of an indicator light according to an embodiment of the present disclosure
  • FIG. 6 shows a schematic diagram of classification and detection of different target objects in an embodiment of the present disclosure
  • Figure 7 shows a schematic view of the structure of a traffic light with multiple bases
  • FIG. 8 shows another flowchart of a method for identifying indication information of an indicator light according to an embodiment of the present disclosure
  • FIG. 9 shows a flowchart of a driving control method according to an embodiment of the present disclosure.
  • Fig. 10 shows a block diagram of a device for identifying indication information of an indicator light according to an embodiment of the present disclosure
  • Fig. 11 shows a block diagram of a driving control device according to an embodiment of the present disclosure
  • Fig. 12 shows a block diagram of an electronic device according to an embodiment of the present disclosure
  • Fig. 13 shows another block diagram of an electronic device according to an embodiment of the present disclosure.
  • the indicator information identification method of the indicator light can be used to perform the indicator information detection of different types of indicator lights.
  • the indicator information identification method of the indicator light can be executed by any electronic device with image processing function, for example Executed by a terminal device or a server or other processing device, where the terminal device may be a user equipment (UE), mobile device, user terminal, terminal, cellular phone, cordless phone, personal digital assistant (PDA) , Handheld devices, computing devices, in-vehicle devices, wearable devices, etc.
  • the indication information recognition method of the indicator light can also be applied to intelligent driving equipment, such as intelligent flying equipment, intelligent vehicles, and blind guide equipment, for intelligent control of the intelligent driving equipment.
  • the method for identifying the indication information of the indicator light can be implemented by the processor invoking a computer-readable instruction stored in the memory.
  • the indicator information identification method of the indicator light provided by the embodiment of the present disclosure can be applied to scenarios such as indicator information identification and detection of indicator lights. For example, it can be used to identify indicator information in application scenarios such as automatic driving and monitoring. This disclosure is not correct. Specific application scenarios are restricted.
  • Fig. 1 shows a flowchart of a method for identifying indicator information of an indicator light according to an embodiment of the present disclosure. As shown in Fig. 1, the method for identifying indicator information of an indicator light includes:
  • the input image may be an image about an indicator light
  • the indicator light may include at least one of a traffic indicator light (for example, a traffic light), an emergency indicator light (for example, a blinking indicator light), and a direction indicator light
  • a traffic indicator light for example, a traffic light
  • an emergency indicator light for example, a blinking indicator light
  • a direction indicator light In other embodiments, it may also be other types of indicator lights.
  • the input image may be an image collected by an image acquisition device, for example, it may be a road driving image collected by an image acquisition device installed in the vehicle, or may also be an image collected by a deployed camera, or in other embodiments
  • the image may also be collected by a handheld terminal device or other device, or the input image may also be an image frame selected from the acquired video stream, which is not specifically limited in the present disclosure.
  • S20 Determine the detection result of the target object based on the input image, the target object includes at least one of the indicator light base and the indicator lamp in the lit state, and the detection result includes the type of the target object and the value of the target area in the input image. position.
  • the target object in the input image can be detected and recognized, and the detection result about the target object can be obtained.
  • the detection result may include the type and location information of the target object.
  • the target detection of the target object in the input image can be realized through the neural network, and the detection result can be obtained.
  • the neural network can be used to detect at least one of the base type of the indicator light, the indicator type of the lighting state, and the position of the base and the position of the lighted indicator in the input image.
  • the detection result of the input image can be obtained through any neural network that can realize the detection of the target object and the classification.
  • the neural network can be a convolutional neural network.
  • the forms of the indicator lights included in the collected input images can be various.
  • traffic indicator lights hereinafter referred to as traffic lights
  • the forms of traffic lights can be various.
  • Figure 2 (a)- Figure 2 (e) the schematic diagrams of the various display states of the traffic light are shown respectively, and Figure 2 (a) shows the different display states of the traffic light.
  • the shape of the lamp base is not limited in this disclosure.
  • the indicator light base may include indicator lights of multiple colors, so the corresponding indicator lights will show multiple states.
  • the traffic light in Figure 2(a) Take the traffic light in Figure 2(a) as an example, where the first group of traffic lights is taken as an example, where L represents the traffic light and D represents the traffic light base.
  • L represents the traffic light
  • D represents the traffic light base.
  • One group is the red, yellow and green lights in the traffic light. All lights are off, and it may be in a fault state at this time.
  • the second group of traffic lights is in the state where the red light is lit
  • the third traffic light is in the state where the yellow light is lit
  • the fourth group of traffic lights is in the state where the green light is lit.
  • the process of identifying the target object it is possible to identify whether the indicator is in the lit state and the color of the indicator in the lit state.
  • the red, yellow, and green texts are only schematically indicating that the traffic lights of the corresponding colors are on.
  • Figure 2(b) shows different arrangements of traffic light bases.
  • traffic lights or other types of indicator lights can be installed on the indicator light base, as shown in Figure 2(b), where the traffic lights are
  • the arrangement on the base can include a horizontal arrangement, a vertical arrangement or a single lamp. Therefore, in the process of identifying the target object, the arrangement of the indicator lights can also be identified.
  • the foregoing is only an exemplary description of the arrangement of the traffic lights on the base. In other embodiments, the arrangement of the indicator lights on the base is also Other types of arrangements can be included.
  • Figure 2(c) shows different application scenarios of traffic lights.
  • traffic lights and other indicators can be set at road intersections, highway intersections, sharp turns, safety warning locations, or travel passages. Therefore, for The identification of the indicator light can also judge and identify the application scenario of the indicator.
  • the actual application scenario in Figure 2(c) is in turn the high-speed intersections marked with "Electronic Toll Collection (ETC)" "Warning signal” and other warning signs such as sudden turns or other dangerous scenes and general scenes.
  • ETC Electronic Toll Collection
  • Figure 2(d) shows a variety of types of traffic lights.
  • the shapes of traffic lights or other indicators are also different according to needs or the needs of the scene.
  • the shapes shown in Figure 2(d) include Arrow-shaped arrow lights, round-spot lights containing round spots, pedestrian lights containing pedestrian signs, or digital lights containing numeric values.
  • various types of lights can also have different colors. This disclosure does not Not limited.
  • Figure 2(e) shows a schematic diagram of combined traffic lights in different situations.
  • the combination of arrow lights in different arrow directions the combination of digital lights and pedestrian lights, which also have indication information such as colors.
  • indication information such as colors.
  • the embodiments of the present disclosure can first detect the target object in the input image to determine the detection result of the target object in the input image, and further obtain the indication information of the target object based on the detection result. For example, by performing target detection on the input image, the type and position of the target object in the input image can be detected, or the detection result can also include the probability of the target object type.
  • classification detection is further performed according to the type of the detected target object to obtain the indication information of the target object, such as light color, value, direction, scene and other information.
  • the embodiment of the present disclosure can divide the type of detection target (ie, target object) into two parts: the indicator light base and the indicator light in the lighting state, where the indicator light in the lighting state can include N types, for example, the type of the indicator light can be It includes at least one of the above-mentioned digital lights, pedestrian lights, arrow lights and round spot lights. Therefore, when the detection of the target object is performed, it can be determined that each target object included in the input image is any one of N+1 types (the base and the N lighting indicators). Alternatively, other types of indicator lights may also be included in other embodiments, which are not specifically limited in the present disclosure.
  • detection may not be performed for the indicator lamp in the off state.
  • the indicator base and the indicator lamp in the on state are not detected, it can be considered that there is no indicator lamp in the input image, so there is no need
  • the process of further identifying the indication information of the target object in S30 is performed.
  • the lamp base is detected but the lamp in the on state is not detected, it can also be regarded as the lamp in the off state. In this case, there is no need to identify the indication information of the target object.
  • S30 Based on the detection result of the target object, identify the target area where the target object is located in the input image to obtain the indication information of the target object.
  • the indication information of the target object when the detection result of the target object is obtained, the indication information of the target object may be further detected, where the indication information is used to describe related attributes of the target object.
  • the indication information of the target object can be used to instruct the smart driving device to generate control instructions based on the indication information. For example, for a target object whose type is a base, at least one of the arrangement of the indicator lights and application scenarios can be identified, and for a target object whose type is a lighted indicator, the light color and arrow of the indicator can be identified At least one of the indication direction, numeric value and other information.
  • the classification and recognition of indication information is performed based on the detection results such as the type of the target object, which helps reduce the recognition complexity in the process of identifying the indication information of the target object. Difficulty in identification, and at the same time, it can easily and conveniently realize the detection and identification of various types of indicator lights in different situations.
  • Fig. 3 shows a flowchart of step S20 in a method for identifying indicator information of an indicator light according to an embodiment of the present disclosure.
  • determining the detection result of the target object based on the input image may include:
  • feature extraction processing may be performed on the input image to obtain image features of the input image.
  • the image features in the input image can be obtained through a feature extraction algorithm, or the image features can be extracted through a neural network trained to achieve feature extraction.
  • a convolutional neural network may be used to obtain image features of an input image, and corresponding image features may be obtained by performing at least one layer of convolution processing on the input image.
  • the convolutional neural network may include a visual geometry group (Visual Geometry Group). , VGG) network, residual network, pyramid feature network at least one, but not as a specific limitation of the present disclosure, image features can also be obtained in other ways.
  • S22 Determine the first position of each candidate area in at least one candidate area of the target object based on the image feature of the input image;
  • the location area of the target object in the input image can be detected according to the image characteristics of the input image, that is, the first position of the candidate area of each target object is obtained.
  • the first position of the candidate area of each target object is obtained for each target object.
  • at least one candidate area can be obtained, and correspondingly, the first position of each candidate area can be obtained.
  • the first position of the embodiment of the present disclosure can be represented by the coordinates of the diagonal vertex position of the candidate area. However, this disclosure does not specifically limit this.
  • the target detection network used to perform target detection may include a base network (base network) module, a regional proposal network (Region Proposal Network, RPN) module, and a classification module.
  • the basic network module is used to perform image processing. Feature extraction processing to obtain image features of the input image.
  • the regional candidate network module is used to detect the candidate region (Region of Interest, ROI) of the target object in the input image based on the image feature of the input image, and the classification module is used to determine the type of the target object in the candidate region based on the image feature of the candidate region The judgment is made to obtain the detection result of the target object in the target area (Box) in the input image.
  • the detection result of the target object includes the type of the target object and the position of the target area.
  • the type of the target object is, for example, a base, an indicator light (such as a round spot light, an arrow light, a pedestrian light, a digital light), Any of the backgrounds.
  • the background can be understood as the image area in the input image except for the area where the base and the lighted indicator are located.
  • the region candidate network can obtain at least one ROI for each target object in the input image, and the ROI with the highest accuracy can be selected through subsequent post-processing.
  • S23 Determine an intermediate detection result of each candidate region based on the image feature at the first position corresponding to each candidate region in the input image, the intermediate detection result includes the prediction type of the target object and the prediction probability that the target object is the prediction type;
  • the type is any one of the indicator lamp base and N kinds of indicator lamps in the lighting state, and N is a positive integer.
  • the type information of the target object in the candidate area can be further classified and recognized, that is, the target object in the candidate area can be obtained
  • the prediction type and the predicted probability for that prediction type can be one of the above N+1 types, for example, it can be any one of a base, a round spot light, an arrow light, a pedestrian light, and a digital light. In other words, it can be predicted whether the type of the target object in the candidate area is a base or one of the N kinds of indicator lights in the lighting state.
  • step S23 may include: for each candidate area, based on the image feature at the first position corresponding to the candidate area, classifying the target object in the candidate area to obtain the target object as each of the at least one preset type.
  • the predicted probability of one preset type wherein, the preset type includes at least one of the indicator base and the N indicator lights in the lighting state, and N is a positive integer; the predicted probability of the at least one preset type is the highest. Set type as the prediction type of the target object in the candidate area, and get the prediction probability of the prediction type.
  • the image feature corresponding to the first position among the image features of the input image may be obtained according to the first position of the candidate region, and the The obtained image feature is determined as the image feature of the candidate area. Further, the prediction probability that the target object in the candidate area is each preset type can be predicted according to the image characteristics of each candidate area.
  • the preset types are the above N+1 types, Such as the base and N types of indicator lights.
  • the preset types may also be N+2 types, and compared to the case of N+1 types, the background type is further included, but the present disclosure does not specifically limit this.
  • the preset type with the highest predicted probability can be determined as the predicted type of the target object in the candidate area, and the corresponding highest predicted probability is The predicted probability of the corresponding prediction type.
  • the image feature of each candidate area may be pooled, so that the scale of the image feature of each candidate area is the same, for example, For each ROI, the size of the image feature can be scaled to 7*7, but this disclosure does not specifically limit this.
  • the image features after the pooling process can be classified to obtain the intermediate detection result corresponding to each candidate frame of each target object.
  • the classification processing on the image features of each candidate region in step S23 may be implemented by using one classifier or multiple classifiers.
  • a classifier to obtain the prediction probability of the candidate area for each preset type
  • N+1 or N+2 classifiers can also be used to detect the prediction probability of the candidate area for each type.
  • the N+1 or N +2 classifiers have a one-to-one correspondence with the preset type, that is, each classifier can be used to obtain the preset result of the corresponding preset type.
  • the convolutional layer when performing classification processing on the candidate region, can also be used to input the image features of the candidate region (or the image features after pooling) into the first convolutional layer to perform convolution processing.
  • the first feature map with dimensions a ⁇ b ⁇ c is obtained, b and c respectively represent the length and width of the first feature map, a represents the number of channels of the first feature map, and the value of a is the total number of preset types (such as N+1), and then perform global pooling processing on the first feature map to obtain a second feature map corresponding to the first feature map.
  • the dimension of the second feature map is a ⁇ d
  • the second feature map is input to In the softmax function
  • a third feature map of dimension a ⁇ d can also be obtained, where d is an integer greater than or equal to 1.
  • d represents the number of columns in the third feature map, for example, it can be 1
  • the corresponding element in the third feature map represents the predicted probability that the target object in the candidate area is each preset type, and each element
  • the corresponding numerical value may be the probability value of the predicted probability, and the order of the probability value corresponds to the order of the set preset type, or each element in the third feature map may be identified by the preset type and the corresponding predicted probability Structure, so as to conveniently determine the correspondence between the preset type and the predicted probability.
  • d may also be another integer value greater than 1, and the predicted probability corresponding to the preset type may be obtained according to the first preset number of elements in the third feature map.
  • the first preset number of columns may be a preset value, for example, it may be 1, but it is not a specific limitation of the present disclosure.
  • the intermediate detection result of each candidate area of each target object can be obtained, and further, the intermediate detection result can be used to obtain the detection result of each target object.
  • S24 Determine the detection result of the target object based on the intermediate detection result of each candidate region in the at least one candidate region and the first position of each candidate region.
  • the intermediate detection results corresponding to all candidate regions for each target object can be obtained, and further, Determine the final detection result of the target object according to the intermediate detection result of each candidate area of the target object, that is, information such as the position and type of the candidate area of the target object.
  • the first position of the candidate region of each target object may be used as the position of the candidate region, or the first position may be optimized to obtain a more accurate first position.
  • the embodiment of the present disclosure may also obtain the position deviation of the corresponding candidate area through the image feature of each candidate area, and adjust the first position of the candidate area according to the position deviation.
  • the image features of the candidate region of each target object can be input to the second convolutional layer to obtain a fourth feature map.
  • the dimension of the fourth feature map is e ⁇ b ⁇ c, where b and c represent the fourth The length and width of the feature map and the third feature map.
  • b and c can also be the length and width of the image feature of the candidate area
  • e represents the channel number of the fourth feature map
  • e can be an integer greater than or equal to 1, for example e can take the value 4.
  • the fifth feature map can be a feature vector of length e, such as e equals 4, and the elements in the fifth feature map are Is the position deviation corresponding to the corresponding candidate area.
  • the dimension of the fifth feature map may be e ⁇ f, and f is a value greater than or equal to 1, indicating the number of columns of the fifth feature map.
  • the preset location area may be a preset location area, such as elements in rows 1-4 and column 1, but it is not a specific limitation of the present disclosure.
  • the first position of the candidate area may be expressed as the horizontal and vertical coordinate values of the two diagonal vertex positions
  • the element in the fifth feature map may be the position offset of the horizontal and vertical coordinate values of the two vertices .
  • the first position of the candidate region can be adjusted according to the corresponding position deviation in the fifth feature map to obtain the first position with higher accuracy.
  • the first convolutional layer and the second convolutional layer are two different convolutional layers.
  • embodiments of the present disclosure can filter out the target area of the target object from the at least one candidate area.
  • the candidate area In the case that only one candidate area is detected for any target object of the input image, it can be determined whether the predicted probability of the predicted type of the target object determined based on the candidate area is greater than the probability threshold. If it is greater than the probability threshold, the candidate can be determined The area is determined as the target area of the target object, and the prediction type corresponding to the candidate area is determined as the type of the target object. If the prediction probability of the prediction type of the target object determined based on the candidate area is less than the probability threshold, the candidate area is discarded, and it is determined that the object in the candidate area does not have any target object to be detected.
  • the intermediate detection result of each candidate area may be based on the intermediate detection result of each candidate area, or the second detection result of each candidate area may be based on the intermediate detection result of each candidate area.
  • a position, the target area is filtered from the multiple candidate areas, the predicted type of the target object in the target area is taken as the type of the target object, and the first position of the target area is taken as the position of the target area where the target object is located to obtain the target The detection result of the object.
  • the step of screening the target area based on the intermediate detection result of the candidate area may include: selecting the candidate area with the highest prediction probability from the multiple candidate areas of the target object, and when the highest prediction probability is greater than the probability threshold In the case of, the first position (or the adjusted first position) of the candidate region corresponding to the highest prediction probability is used as the target region of the target object, and the prediction type corresponding to the highest prediction probability is determined as the target object type .
  • the step of filtering out the target area of the target object based on the first position of the candidate area may include: using a non-maximum suppression algorithm (Non-maximum suppression, NMS) to select the target of the target object from multiple candidate areas area.
  • NMS non-maximum suppression
  • the candidate area with the largest prediction probability is selected from the multiple candidate areas of the target object in the input image, which is hereinafter referred to as the first candidate area.
  • the overlap area value Intersection over Union, IOU
  • any candidate area is discarded. If after comparing the IOUs, the remaining candidate areas are discarded, the first candidate area is the target area of the target object, and the predicted type of the target object obtained based on the first candidate area may be the type of the target object.
  • the candidate area with the highest predicted probability in the second candidate area can be re-used as the new first candidate area, Continue to obtain the remaining candidate regions in the second candidate region and the IOU of the new first candidate region, and also discard the second candidate region whose IOU is greater than the area threshold until there is no difference between the first candidate region (or the new candidate region) Candidate regions whose IOU is greater than the area threshold.
  • Each first candidate area obtained in the foregoing manner can be determined as a target area of each target object.
  • candidate regions with a predicted probability greater than the probability threshold can also be selected from the candidate regions of each target object through the probability threshold, and then the target region of each target object is obtained through the above-mentioned NMS algorithm. At the same time, the prediction type for the target object in the target area is obtained, that is, the detection result of the target object is determined.
  • the detection result of the target object existing in the input image can be obtained, that is, the type and corresponding position of the target object can be easily determined.
  • a detection frame can be obtained for each target object (such as a lighted indicator lamp, indicator base)
  • the detection result can include the input image
  • the position of the indicator in the lit state and the type of the indicator for example, the detection result can be expressed as (x1, y1, x2, y2, label1, score1).
  • (x1, y1), (x2, y2) are the position coordinates of the target area of the indicator lamp in the lit state (the coordinates of the two diagonal points), and label1 represents the type identification of the indicator lamp in the lit state (1 to One of N+1, such as 2, can be represented as a digital light), and score1 represents the confidence of the detection result (ie, the predicted probability).
  • the test result is expressed as (x3, y3, x4, y4, label2, scor12).
  • (x3, y3), (x4, y4) are the position coordinates of the target area of the base (the coordinates of the two diagonal points)
  • label2 represents the type identification of the base (one of 1 to N, such as 1)
  • score2 Indicates the confidence level of the test result.
  • the identifier of the base may be 1, and the remaining N identifiers may be N types of indicator lights in the lighting state. In some possible implementations, N+2 may also be identified to indicate the target area of the background. No specific restrictions on this
  • the detection result for the target object can be obtained simply and conveniently.
  • the detection result already includes the indicator light or the type information of the base, the classification pressure of the subsequent classifier can be reduced.
  • the detection result of the target object in the input image when the detection result of the target object in the input image is obtained, it may be further determined based on the detection result whether the indicator light is malfunctioning, or information such as the collection environment of the input image is collected.
  • the type of the detected target object in the result of the target object of the input image, only includes the indicator lamp base, and does not include any type of indicator lamp in the lighting state.
  • the fault information is sent to the server or other management equipment, and the fault information may include the fault condition that the indicator light is off, and the location information of the fault light (determined based on the foregoing collection location).
  • the acquisition can be determined at this time
  • the collection environment of the input image is a dark environment or a dark state, where the dark state or the dark environment refers to an environment where the light brightness is less than the preset brightness.
  • the preset brightness can be set according to different locations or different weather conditions. There is no specific limitation.
  • FIG. 5 shows a flowchart of step S30 in a method for identifying indicator information of an indicator light according to an embodiment of the present disclosure, in which, based on the detection result of the target object, the target area in the input image is identified to obtain the target
  • the indication information of the object may include:
  • S31 Determine a classifier matching the target object based on the target object type in the detection result of the target object;
  • S32 Use the matched classifier to recognize the image feature of the target area in the input image to obtain the indication information of the target object.
  • the classifiers matching the target object include at least one type, and each classifier may correspond to one or more types of target objects.
  • the classification detection of the indication information can be performed, such as the scene information in the base, the arrangement of the indicator lights, the color of the indicator lights, the description, and the indication direction. Classification and identification of at least one type of information.
  • the embodiments of the present disclosure may use different classifiers to perform classification and recognition of different indication information, so the classifier that performs classification and recognition may be determined first.
  • FIG. 6 shows a schematic diagram of classification and detection of different target objects in an embodiment of the present disclosure.
  • the classification and recognition of the indicator information can be further performed on the target object of the base type to obtain the arrangement of the indicator light and the scene where the indicator light is located.
  • the arrangement can include horizontal arrangement, vertical arrangement, and arrangement of a single indicator.
  • the scenes may include highway intersections, sharp turns, general scenes, etc. The above are only exemplary descriptions of arrangement modes and scenes, and may also include other arrangement modes or scenes, which are not specifically limited in the present disclosure.
  • the light color of the round-spot light can be classified and identified to obtain the light color (such as red, Green, yellow) instructions.
  • the type of the identified target object is a digital indicator light in the lit state
  • the numerical value such as 1, 2, 3, etc.
  • the light color can be classified and identified to obtain the light color and the numerical indicator information.
  • the direction of the arrow such as forward, left, right, etc.
  • the light color can be classified and identified to obtain the light color And instructions for directions.
  • the type of the identified target object is an indicator lamp (pedestrian lamp) for pedestrian identification
  • the lighting color can be recognized to obtain the indication information of the lighting color.
  • the embodiments of the present disclosure can perform identification of different indication information for different types of target objects in the detection result of the target object, so that the indication information of the indicator light can be obtained more conveniently and accurately.
  • the image feature corresponding to the target area where the target object of the corresponding type is located can be input into the matching classifier to obtain the classification result, that is, to obtain the corresponding indication information.
  • the determined matching classifier includes at least one of a first classifier and a second classifier , Where the first classifier is used to classify and recognize the arrangement of the indicator lights in the base, and the second classifier is used to classify and identify the scene where the indicator lights are located.
  • the scene of the indicator light can be obtained, for example, the scene information can be obtained by means of text recognition.
  • determining the matching classifier includes a classifier for performing color attributes of the round-spot light or pedestrian light. Recognized third classifier. At this time, the image features of the target area corresponding to the target object of the round spot light type or the pedestrian light type can be input into the matching third classifier to obtain the color attribute of the indicator light.
  • the classifier that determines the match includes a fourth classifier used to identify the color attribute of the arrow light and the direction attribute.
  • the fifth classifier used to identify the color attribute of the arrow light and the direction attribute.
  • the image features of the target area corresponding to the target object of the arrow light type can be input into the matching fourth and fifth classifiers, and the fourth and fifth classifiers can be used to compare the target area where the target object is located. Recognize the image characteristics of the arrow, and get the color attribute of the arrow light and the direction attribute of the arrow light.
  • the classifier that determines the match includes the sixth classifier for the color attribute of the digital lamp and the The seventh classifier for identifying numerical attributes.
  • the image features of the target area corresponding to the target object of the digital light type can be input into the matching sixth classifier and seventh classifier, and based on the sixth classifier and seventh classifier, the target area where the target object is located Recognize the image features of the digital lights, and obtain the color attributes and numerical direction attributes of the digital lights.
  • the third, fourth, and sixth classifiers that perform the classification and recognition of color attributes may be the same classifier or different classifiers, and the present disclosure does not specifically limit this .
  • the above-mentioned method for acquiring the image feature of the target area may include determining the image feature of the target area based on the image feature of the input image obtained by feature extraction of the input image and the location information of the target area. That is to say, the feature corresponding to the position information of the target area can be directly obtained from the image feature of the input image as the image feature of the target area. Alternatively, a sub-image corresponding to the target area in the input image may be obtained, and then feature extraction is performed on the sub-image, such as convolution processing, to obtain the image characteristics of the sub-image, thereby determining the image characteristics of the target area.
  • the image characteristics of the target area may also be obtained in other ways, which is not specifically limited in the present disclosure.
  • the indication information of the target object in each target area can be obtained.
  • the detection of different indication information can be performed by different classifiers to make the classification result more accurate.
  • the matching classifier is further used for classification and recognition, instead of using all the classifiers for identification, the classifier resources can be effectively used and the classification speed can be accelerated.
  • the input image may include multiple indicator bases and multiple indicator lights in a lighting state.
  • FIG. 7 shows a schematic diagram of the structure of a traffic light with multiple bases.
  • the base and the lighting state indicator lights can be matched.
  • there are two indicator lamps D1 and D2 and each indicator lamp base can include a corresponding indicator lamp.
  • L1, L2, and L3 By matching the indicator base and the indicator lamp in the lighting state, it can be determined that the indicator lamp L1 in the lighting state matches the indicator lamp base D1, and the indicators L2 and L3 match the base D2.
  • FIG. 8 shows another flow chart of a method for identifying indicator information of an indicator light according to an embodiment of the present disclosure, wherein the method for identifying indicator information of indicator light further includes a matching process between the indicator light base and the indicator light in the lighting state, Specifically:
  • the first indicator light base determines an indicator light that matches the first indicator light base in a lighting state; the first indicator light base is one of at least two indicator light bases;
  • the obtained detection result of the target object may include the first position of the target area of the target object of the base type and the second position of the target area where the indicator lamp is on.
  • the embodiment of the present disclosure may be based on the The first position and the second position of each indicator light determine whether the base and the lighted indicator match.
  • the first area that intersects between the target area where the at least one indicator light is located and the target area where the first indicator light base is located may be determined, And, determining the second area of the target area where at least one lighted indicator is located; in response to the first area corresponding to the first indicator light in the lighted state, and the second area of the first indicator light in the lighted state
  • the ratio between is greater than the set area threshold, and it is determined that the first indicator lamp in the lighting state matches the first indicator lamp base; wherein the first indicator lamp in the lighting state is one of at least one indicator lamp in the lighting state .
  • each base and each base can be determined according to the first position of the target area of the first indicator light base and the second position of the target area of each lighted indicator The first area S1 that intersects or overlaps between the target areas of the two indicator lights. If there is a light-on indicator light (the first indicator light) and the indicator light base, the first area S1 and the lighted state indicator If the ratio (S1/S2) between the second area S2 of the target area of the lamp is greater than the area threshold, it can be determined that the first indicator light matches the first indicator light base.
  • the multiple first indicators can be used as indicator lights matching the first indicator base at the same time, or the first indicator with the largest ratio can be used
  • the light is determined to be an indicator light in a lighting state matching the first indicator light base.
  • the preset number of indicator lights with the largest S1/S2 ratio between the first indicator light base and the first indicator light base may be determined as the indicator lights matching the first indicator light base.
  • the preset number can be 2, but it is not a specific limitation of the present disclosure.
  • the area threshold may be a preset value, such as 0.8, but it is not a specific limitation of the present disclosure.
  • the indication information obtained by the indicator base and the matching indicator light in the lighting state can be combined to obtain the indicator information.
  • the indicator base D1 and the indicator light L1 in the lighting state can be combined.
  • the determined indicator information includes that the scene is a general scene, the indicator lights are arranged horizontally, and the indicator lights in the lighting state It is a round spot light and the color is red.
  • the indicator base D2 and the indicator lights L2 and L3 of the lighting state can be combined.
  • the determined indicator information includes the scene as a general scene, the arrangement of the indicators is horizontal, and the indicator lights in the lit state are arrows. Lights, and include a right arrow light and a forward arrow light, wherein the color of the right arrow light is red, and the color of the forward arrow light is green.
  • the base may be determined to be in an off state. It can be determined that the indicator light corresponding to the base is a fault light. For the indicator lamp of the lighting state that cannot find the matching indicator base, the indicator information corresponding to the indicator lamp of the lighting state is output separately. This situation is often caused by the inconspicuous visual features of the base, for example, it is difficult to detect the base at night.
  • the obtained input image may be an image of the front or rear of the vehicle collected in real time.
  • the driving may be further generated based on the obtained indication information.
  • the driving parameters may include driving conditions such as driving speed, driving direction, control mode, and stopping.
  • the algorithm model used in the embodiments of the present disclosure may include two parts, one part is the target detection network for performing target detection shown in FIG. 4, and the other part is the classification network for performing classification and recognition of indication information.
  • the target detection network may include a base network (base network) module, a regional candidate network (RPN) module, and a classification module.
  • the basic network module is used to perform feature extraction processing of the input image to obtain the input image Image characteristics.
  • the regional candidate network module is used to detect the candidate region (ROI) of the target object in the input image based on the image feature of the input image, and the classification module is used to judge the type of the target object in the candidate region based on the image feature of the candidate region, and get The detection result of the target object of the input image.
  • ROI candidate region
  • the classification module is used to judge the type of the target object in the candidate region based on the image feature of the candidate region, and get The detection result of the target object of the input image.
  • the input of the target detection network is the input image
  • the output is the 2D detection frame of a number of target objects (that is, the target area of the target object)
  • each detection frame can be expressed as (x1, y1, x2, y2, label, score) .
  • x1, y1, x2, y2 are the position coordinates of the detection frame
  • label is the category (the value range is 1 to N+1, the first category represents the base, and the other categories represent various lighting status indicators.
  • the process of target detection may include: inputting the input image to Base Network to obtain the image characteristics of the input image.
  • a regional candidate network (Region Proposal Network, RPN) is used to generate a candidate frame ROI (Region of interest) of the indicator, which includes the candidate frame of the base and the candidate frame of the indicator light in the lit state.
  • the pooling layer can be used to obtain the feature map of the candidate frame with a fixed size. For example, for each ROI, the size of the feature map is scaled to 7*7, and then the classification module is used to determine the category of N+2 (the background category is added) to obtain the candidate frame of each target object in the input image The forecast type and location. Then, post-processing such as NMS and threshold is performed to obtain the final detection frame of the target object (the candidate frame corresponding to the target area).
  • the embodiment of the present disclosure divides the indicator lights in the lighting state into N types of rationality among the detected target objects:
  • the lights in the lighting state are subdivided into N different categories, which is convenient for adjusting the parameters of the model and adjusting and optimizing them separately.
  • the indication information of the target object can be further identified.
  • the instruction information can be classified and identified by a matching classifier.
  • a classification module including a plurality of classifiers can be used to perform the identification of the indication information of the target object.
  • the classification module may include multiple types of classifiers for performing classification and recognition of different indication information, or may also include a convolutional layer for extracting features, which is not specifically limited in the present disclosure.
  • the input of the classification module may be the detected image feature corresponding to the target area of the target object, and the output may be the indication information corresponding to the target object in the target area.
  • the specific process may include: inputting the detection frame of the target area of the target object, selecting a classifier matching the target object type (1 to N+1) in the detection frame, and obtaining the corresponding classification result. If it is the detection frame of the indicator base, since the indicator base can be regarded as a simple whole, all the classifiers of the indicator base are activated, for example, the classifiers used to identify the scene and the arrangement are all activated to identify the scene Attributes and arrangement attributes; if it is the detection frame of the indicator light in the lit state, different types of indicator lights in the lit state need to select different classifiers, for example, the arrow lights correspond to the "color" and "arrow direction”. A classifier, the round spotlight corresponds to the "color" classifier, and so on. In addition, if other attribute determination requirements are added, other classifiers can also be added, which is not specifically limited in the present disclosure.
  • the embodiments of the present disclosure may first perform target detection processing on the input image to obtain the target object detection result, where the target object detection result may include information such as the position and type of the target object, and then further based on the target object detection result Perform the identification of the instructions of the target object.
  • the present disclosure divides the detection process of the target object into two detection processes of the base and the indicator light in the lighting state, and realizes the first discrimination of the target object during the detection process, and then further recognizes the target object based on the detection result of the target object. It is helpful to reduce the recognition complexity in the process of recognizing the indication information of the target object, reduce the recognition difficulty, and can simply and conveniently realize the detection and recognition of various types of indicator lights in different situations.
  • the embodiment of the present disclosure only uses picture information without using other sensors to realize indicator light detection and indication information determination. At the same time, the embodiment of the present disclosure can detect different types of indicator lights, which has better applicability.
  • Fig. 9 shows a flowchart of a driving control method according to an embodiment of the present disclosure.
  • the driving control method can be applied to devices such as smart vehicles, smart aircrafts, toys, etc., which can adjust driving parameters according to control instructions.
  • the driving control method may include:
  • S100 Use the image acquisition device in the intelligent driving device to collect driving images
  • the image acquisition device provided in the smart driving device may collect driving images, or may also receive form images at the driving position collected by other devices.
  • S200 Perform the indicator light indication information identification method on the driving image to obtain the indication information for the driving image;
  • the detection processing of the indication information is performed on the driving image, that is, the method for identifying the indication information of the indicator light described in the above embodiment is executed to obtain the indication information of the indicator light in the driving image.
  • S300 Use the instruction information to generate a control instruction of the intelligent driving device.
  • the driving parameters of the driving device can be controlled in real time. That is, the control instruction for controlling the functional driving device can be generated according to the obtained instruction information.
  • the control instruction can be used to control the driving parameters of the intelligent driving device. It includes at least one of driving speed, driving direction, driving mode, or driving state.
  • the parameter control or control instruction type of the driving device those skilled in the art can set according to the existing technical means and requirements, which is not specifically limited in the present disclosure.
  • intelligent control of the intelligent driving device can be realized. Since the acquisition process of the instruction information is simple, fast, and high in accuracy, the control efficiency and accuracy of the driving device can be improved.
  • the writing order of the steps does not mean a strict execution order but constitutes any limitation on the implementation process.
  • the specific execution order of each step should be based on its function and possibility.
  • the inner logic is determined. Without violating logic, the different implementation manners provided by the present disclosure can be combined with each other.
  • the present disclosure also provides an indicator light indication information recognition device, a driving control device, an electronic device, a computer-readable storage medium, and a program, all of which can be used to implement any indicator light indication information recognition method and / Or driving control methods, corresponding technical solutions and descriptions and refer to the corresponding records in the method section, and will not be repeated.
  • FIG. 10 shows a block diagram of a device for identifying indication information of an indicator light according to an embodiment of the present disclosure.
  • the device for identifying indication information of an indicator light includes:
  • the obtaining module 10 is used to obtain an input image
  • the detection module 20 is configured to determine a detection result of a target object based on the input image, the target object includes at least one of an indicator lamp base and an indicator lamp in a lighting state, and the detection result includes the target object The type of and the position of the target area where the target object is located in the input image;
  • the recognition module 30 is configured to recognize the target area where the target object is located in the input image based on the detection result of the target object to obtain the indication information of the target object.
  • the determining module is further used for:
  • the intermediate detection result of each candidate area including the prediction type of the target object and the target object being The prediction probability of the prediction type;
  • the prediction type is any one of the indicator lamp base and the N types of indicator lamps in the lighting state, and N is a positive integer;
  • the detection result of the target object is determined based on the intermediate detection result of each candidate region in the at least one candidate region and the first position of each candidate region.
  • the determining module is further configured to: for each candidate area, classify the target object in the candidate area based on the image feature at the first position corresponding to the candidate area to obtain The target object is the predicted probability of each preset type in the at least one preset type; wherein, the preset type includes at least one of an indicator light base and N types of indicator lights in the lighting state, N Is a positive integer;
  • the preset type with the highest prediction probability among the at least one preset type is used as the prediction type of the target object in the candidate area, and the prediction probability of the prediction type is obtained.
  • the determining module is further configured to: before determining the detection result of the target object based on the intermediate detection result of each candidate region in the at least one candidate region and the first position of each candidate region , Determining the position deviation of the first position of each candidate region based on the image feature of the input image;
  • the determining module is further configured to, when there are at least two candidate regions of the target object, based on the intermediate detection result of each of the at least two candidate regions, or based on The intermediate detection result of each candidate area and the first position of each candidate area, selecting the target area from the at least two candidate areas;
  • the predicted type of the target object in the target area is taken as the type of the target object
  • the first position of the target area is taken as the position of the target area where the target object is located
  • the detection result of the target object is obtained .
  • the determining module is further configured to determine that the indicator light is in a fault state when the detection result of the target object includes only the detection result corresponding to the indicator light base;
  • the detection result of the target object only includes the detection result corresponding to the indicator lamp in the lit state, it is determined that the state of the scene in which the input image is collected is the dark state.
  • the recognition module is further configured to determine a classifier matching the target object based on the type of the target object in the detection result of the target object;
  • the matching classifier is used to identify the image features of the target area in the input image to obtain the indication information of the target object.
  • the identification module is further configured to determine that the matching classifier includes an arrangement for arranging the indicators in the indicator base when the type of the target object is the indicator base.
  • the first classifier that recognizes by using the first classifier, uses the first classifier to recognize the image feature of the target area where the target object is located, and determines the arrangement of the indicator lights in the indicator light base; and/or,
  • the classifier that determines the match includes a second classifier for recognizing the scene where the indicator is located, and using the second classifier to recognize the image feature of the target area where the target object is located, and determine the indicator Information about the scene where the light is located.
  • the recognition module is further configured to determine that the matching classifier includes a classifier used to identify the color attribute of the round spot light when the type of the target object is a round spot light or a pedestrian light.
  • the third classifier is further configured to determine that the matching classifier includes a classifier used to identify the color attribute of the round spot light when the type of the target object is a round spot light or a pedestrian light.
  • the third classifier is used to identify the image feature of the target area where the target object is located, and determine the color attribute of the round spot light or the pedestrian light.
  • the recognition module is further configured to determine that the matching classifier includes a fourth classifier and direction for determining the color attribute of the arrow lamp when the type of the target object is an arrow light.
  • the image features of the target area where the target object is located are recognized, and the color attribute and the direction attribute of the arrow light are respectively determined.
  • the recognition module is further configured to determine that the matching classifier includes a sixth classifier and a numerical value for the color attribute of the digital lamp when the type of the target object is a digital lamp.
  • the image features of the target area where the target object is located are recognized, and the color attribute and the numerical attribute of the digital lamp are determined respectively.
  • the device further includes a matching module, which is used to determine a match with the first indicator base for the first indicator base when the input image includes at least two indicator bases.
  • the matching module is also used to:
  • the ratio of the first area between the first indicator lamp in the lit state and the first indicator lamp base to the second area of the first indicator lamp in the lit state is greater than the set area In the case of a threshold, it is determined that the first indicator light in the lighting state matches the first indicator light base;
  • the first indicator lamp in the lighting state is one of the at least one indicator lamp in the lighting state.
  • FIG. 11 shows a block diagram of a driving control device according to an embodiment of the present disclosure.
  • the driving control device includes:
  • An image acquisition module 100 which is set in an intelligent driving device and used to collect driving images of the intelligent driving device;
  • the image processing module 200 is configured to execute the indicator light indication information recognition method described in any one of the first aspect on the driving image to obtain the indication information for the driving image;
  • the control module 300 is configured to use the instruction information to generate a control instruction for controlling the intelligent driving device.
  • the functions or modules contained in the device provided in the embodiments of the present disclosure can be used to execute the methods described in the above method embodiments.
  • the functions or modules contained in the device provided in the embodiments of the present disclosure can be used to execute the methods described in the above method embodiments.
  • the embodiments of the present disclosure also provide a computer-readable storage medium on which computer program instructions are stored, and the computer program instructions implement the above-mentioned method when executed by a processor.
  • the computer-readable storage medium may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium.
  • An embodiment of the present disclosure also provides an electronic device, including: a processor; a memory for storing executable instructions of the processor; wherein the processor is configured as the above method.
  • the embodiment of the present disclosure also proposes a computer program, including computer readable code, when the computer readable code is executed in an electronic device, the processor in the electronic device is executed to implement the above method.
  • the electronic device can be provided as a terminal, server or other form of device.
  • Fig. 12 shows a block diagram of an electronic device according to an embodiment of the present disclosure.
  • the electronic device 800 may be a mobile phone, a computer, a digital broadcasting terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, and other terminals.
  • the electronic device 800 may include one or more of the following components: processing component 802, memory 804, power supply component 806, multimedia component 808, audio component 810, input/output (I/O) interface 812, sensor component 814 , And communication component 816.
  • the processing component 802 generally controls the overall operations of the electronic device 800, such as operations associated with display, telephone calls, data communications, camera operations, and recording operations.
  • the processing component 802 may include one or more processors 820 to execute instructions to complete all or part of the steps of the foregoing method.
  • the processing component 802 may include one or more modules to facilitate the interaction between the processing component 802 and other components.
  • the processing component 802 may include a multimedia module to facilitate the interaction between the multimedia component 808 and the processing component 802.
  • the memory 804 is configured to store various types of data to support operations in the electronic device 800. Examples of these data include instructions for any application or method operated on the electronic device 800, contact data, phone book data, messages, pictures, videos, etc.
  • the memory 804 can be implemented by any type of volatile or nonvolatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable Programmable Read Only Memory (EPROM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), Magnetic Memory, Flash Memory, Magnetic Disk or Optical Disk.
  • SRAM static random access memory
  • EEPROM electrically erasable programmable read-only memory
  • EPROM erasable Programmable Read Only Memory
  • PROM Programmable Read Only Memory
  • ROM Read Only Memory
  • Magnetic Memory Flash Memory
  • Magnetic Disk Magnetic Disk or Optical Disk.
  • the power supply component 806 provides power for various components of the electronic device 800.
  • the power supply component 806 may include a power management system, one or more power supplies, and other components associated with the generation, management, and distribution of power for the electronic device 800.
  • the multimedia component 808 includes a screen that provides an output interface between the electronic device 800 and the user.
  • the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user.
  • the touch panel includes one or more touch sensors to sense touch, sliding, and gestures on the touch panel. The touch sensor may not only sense the boundary of a touch or slide action, but also detect the duration and pressure related to the touch or slide operation.
  • the multimedia component 808 includes a front camera and/or a rear camera. When the electronic device 800 is in an operation mode, such as a shooting mode or a video mode, the front camera and/or the rear camera can receive external multimedia data. Each front camera and rear camera can be a fixed optical lens system or have focal length and optical zoom capabilities.
  • the audio component 810 is configured to output and/or input audio signals.
  • the audio component 810 includes a microphone (MIC).
  • the microphone is configured to receive external audio signals.
  • the received audio signal may be further stored in the memory 804 or transmitted via the communication component 816.
  • the audio component 810 further includes a speaker for outputting audio signals.
  • the I/O interface 812 provides an interface between the processing component 802 and a peripheral interface module.
  • the peripheral interface module may be a keyboard, a click wheel, a button, and the like. These buttons may include but are not limited to: home button, volume button, start button, and lock button.
  • the sensor component 814 includes one or more sensors for providing the electronic device 800 with various aspects of state evaluation.
  • the sensor component 814 can detect the on/off status of the electronic device 800 and the relative positioning of the components.
  • the component is the display and the keypad of the electronic device 800.
  • the sensor component 814 can also detect the electronic device 800 or the electronic device 800.
  • the position of the component changes, the presence or absence of contact between the user and the electronic device 800, the orientation or acceleration/deceleration of the electronic device 800, and the temperature change of the electronic device 800.
  • the sensor component 814 may include a proximity sensor configured to detect the presence of nearby objects when there is no physical contact.
  • the sensor component 814 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications.
  • the sensor component 814 may also include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor or a temperature sensor.
  • the communication component 816 is configured to facilitate wired or wireless communication between the electronic device 800 and other devices.
  • the electronic device 800 can access a wireless network based on a communication standard, such as WiFi, 2G, or 3G, or a combination thereof.
  • the communication component 816 receives a broadcast signal or broadcast related information from an external broadcast management system via a broadcast channel.
  • the communication component 816 further includes a near field communication (NFC) module to facilitate short-range communication.
  • the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.
  • RFID radio frequency identification
  • IrDA infrared data association
  • UWB ultra-wideband
  • Bluetooth Bluetooth
  • the electronic device 800 can be implemented by one or more application specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field A programmable gate array (FPGA), controller, microcontroller, microprocessor, or other electronic components are implemented to implement the above methods.
  • ASIC application specific integrated circuits
  • DSP digital signal processors
  • DSPD digital signal processing devices
  • PLD programmable logic devices
  • FPGA field A programmable gate array
  • controller microcontroller, microprocessor, or other electronic components are implemented to implement the above methods.
  • a non-volatile computer-readable storage medium or a volatile computer-readable storage medium is also provided, such as the memory 804 including computer program instructions, which can be processed by the electronic device 800.
  • the device 820 executes to complete the above method.
  • Fig. 13 shows another block diagram of an electronic device according to an embodiment of the present disclosure.
  • the electronic device 1900 may be provided as a server. 13
  • the electronic device 1900 includes a processing component 1922, which further includes one or more processors, and a memory resource represented by a memory 1932 for storing instructions executable by the processing component 1922, such as application programs.
  • the application program stored in the memory 1932 may include one or more modules each corresponding to a set of instructions.
  • the processing component 1922 is configured to execute instructions to perform the above-described methods.
  • the electronic device 1900 may also include a power supply component 1926 configured to perform power management of the electronic device 1900, a wired or wireless network interface 1950 configured to connect the electronic device 1900 to a network, and an input output (I/O) interface 1958 .
  • the electronic device 1900 can operate based on an operating system stored in the memory 1932, such as Windows ServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM or the like.
  • a non-volatile computer-readable storage medium or a volatile computer-readable storage medium is also provided, such as the memory 1932 including computer program instructions, which can be processed by the electronic device 1900.
  • the component 1922 executes to complete the above method.
  • the present disclosure may be a system, method, and/or computer program product.
  • the computer program product may include a computer-readable storage medium loaded with computer-readable program instructions for enabling a processor to implement various aspects of the present disclosure.
  • the computer-readable storage medium may be a tangible device that can hold and store instructions used by the instruction execution device.
  • the computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing.
  • Computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM) Or flash memory), static random access memory (SRAM), portable compact disk read-only memory (CD-ROM), digital versatile disk (DVD), memory stick, floppy disk, mechanical encoding device, such as a printer with instructions stored thereon
  • RAM random access memory
  • ROM read-only memory
  • EPROM erasable programmable read-only memory
  • flash memory flash memory
  • SRAM static random access memory
  • CD-ROM compact disk read-only memory
  • DVD digital versatile disk
  • memory stick floppy disk
  • mechanical encoding device such as a printer with instructions stored thereon
  • the computer-readable storage medium used herein is not interpreted as a transient signal itself, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission medium (for example, light pulses through fiber optic cables), or through wires Transmission of electrical signals.
  • the computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to various computing/processing devices, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and/or a wireless network.
  • the network may include copper transmission cables, optical fiber transmission, wireless transmission, routers, firewalls, switches, gateway computers, and/or edge servers.
  • the network adapter card or network interface in each computing/processing device receives computer-readable program instructions from the network, and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing/processing device .
  • the computer program instructions used to perform the operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, status setting data, or in one or more programming languages.
  • Source code or object code written in any combination, the programming language includes object-oriented programming languages such as Smalltalk, C++, etc., and conventional procedural programming languages such as "C" language or similar programming languages.
  • Computer-readable program instructions can be executed entirely on the user's computer, partly on the user's computer, executed as a stand-alone software package, partly on the user's computer and partly executed on a remote computer, or entirely on the remote computer or server carried out.
  • the remote computer can be connected to the user's computer through any kind of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (for example, using an Internet service provider to access connection).
  • LAN local area network
  • WAN wide area network
  • an electronic circuit such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), can be customized by using the status information of the computer-readable program instructions.
  • the computer-readable program instructions are executed to realize various aspects of the present disclosure.
  • These computer-readable program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine such that when these instructions are executed by the processor of the computer or other programmable data processing device , A device that implements the functions/actions specified in one or more blocks in the flowchart and/or block diagram is produced. It is also possible to store these computer-readable program instructions in a computer-readable storage medium. These instructions make computers, programmable data processing apparatuses, and/or other devices work in a specific manner, so that the computer-readable medium storing instructions includes An article of manufacture, which includes instructions for implementing various aspects of the functions/actions specified in one or more blocks in the flowchart and/or block diagram.
  • each block in the flowchart or block diagram may represent a module, program segment, or part of an instruction, and the module, program segment, or part of an instruction contains one or more functions for implementing the specified logical function.
  • Executable instructions may also occur in a different order from the order marked in the drawings. For example, two consecutive blocks can actually be executed in parallel, or they can sometimes be executed in the reverse order, depending on the functions involved.
  • each block in the block diagram and/or flowchart, and the combination of the blocks in the block diagram and/or flowchart can be implemented by a dedicated hardware-based system that performs the specified functions or actions Or it can be realized by a combination of dedicated hardware and computer instructions.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Multimedia (AREA)
  • Artificial Intelligence (AREA)
  • Medical Informatics (AREA)
  • General Health & Medical Sciences (AREA)
  • Databases & Information Systems (AREA)
  • Software Systems (AREA)
  • Computing Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Atmospheric Sciences (AREA)
  • Image Analysis (AREA)
  • Traffic Control Systems (AREA)

Abstract

一种指示灯的指示信息识别方法及装置、电子设备和存储介质，方法包括：获取输入图像(S10)；基于输入图像，确定目标对象的检测结果，目标对象包括指示灯底座、点亮状态的指示灯中的至少一种，检测结果包括目标对象的类型、输入图像中目标对象所在的目标区域的位置(S20)；基于目标对象的检测结果，对输入图像中目标对象所在的目标区域进行识别，得到目标对象的指示信息(S30)。

Description

指示灯的指示信息识别方法及装置、电子设备和存储介质
本公开要求在2019年6月27日提交中国专利局、申请号为201910569896.8、申请名称为“指示灯的指示信息识别方法及装置、电子设备和存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本公开中。
技术领域
本公开涉及计算机视觉技术领域，尤其涉及一种指示灯的指示信息识别方法及装置、电子设备和存储介质。
背景技术
交通灯是设置在马路上，为车辆和行人提供指导信号的装置。道路路况是非常复杂的，随时可能会发生突变情况或者意外，交通灯可以通过调节不同对象的通过时间，来调节诸多矛盾以及预防意外的发生。例如在交叉路口，不同车道的车辆会抢占通过路口，从而造成矛盾。
在实际应用中，交通灯可以应用在不同的场景、具有不同的形状和类型，其中具有复杂的承接关系。
发明内容
本公开提出了一种指示灯的指示信息识别的技术方案。
根据本公开的一方面，提供了一种指示灯的指示信息识别方法，其包括：
获取输入图像；
基于所述输入图像，确定目标对象的检测结果，所述目标对象包括指示灯底座、点亮状态的指示灯中的至少一种，所述检测结果包括所述目标对象的类型、所述输入图像中所述目标对象所在的目标区域的位置；
基于所述目标对象的检测结果，对所述输入图像中所述目标对象所在的目标区域进行识别，得到所述目标对象的指示信息。
在一些可能的实施方式中，所述基于所述输入图像，确定目标对象的检测结果，包括：
提取所述输入图像的图像特征；
基于所述输入图像的图像特征，确定所述目标对象的至少一个候选区域中每个候选区域的第一位置；
基于所述输入图像中每个候选区域对应的第一位置处的图像特征，确定每个候选区域的中间检测结果，所述中间检测结果包括所述目标对象的预测类型和所述目标对象为所述预测类型的预测概率；所述预测类型为指示灯底座、N种点亮状态的指示灯中的任一种，N为正整数；
基于至少一个候选区域中每个候选区域的中间检测结果以及每个候选区域的第一位置，确定所述目标对象的检测结果。
在一些可能的实施方式中，基于所述输入图像中每个候选区域对应的第一位置处的图像特征，确定每个候选区域的中间检测结果，包括：
针对每个候选区域，基于该候选区域对应的第一位置处的图像特征，对该候选区域内的所述目标对象进行分类，得到所述目标对象为所述至少一种预设类型中每种预设类型的预测概率；其中，所述预设类型包括指示灯底座、N种点亮状态的指示灯中的至少一种，N为正整数；
将所述至少一种预设类型中预测概率最高的预设类型，作为该候选区域内所述目标对象的预测类型，并得到所述预测类型的预测概率。
在一些可能的实施方式中，在基于至少一个候选区域中每个候选区域的中间检测结果以及每个候选区域的第一位置，确定所述目标对象的检测结果之前，还包括：
基于所述输入图像的图像特征，确定每个候选区域的第一位置的位置偏差；
利用每个候选区域对应的位置偏差，调整每个候选区域的第一位置。
在一些可能的实施方式中，所述基于至少一个候选区域中每个候选区域的中间检测结果以及每个候选区域的第一位置，确定所述目标对象的检测结果，包括：
响应于所述目标对象的候选区域为至少两个，基于至少两个候选区域中每个候选区域的中间检测 结果，或者，基于每个候选区域的中间检测结果以及每个候选区域的第一位置，从所述至少两个候选区域中筛选出目标区域；
将所述目标区域内所述目标对象的预测类型作为所述目标对象的类型，将所述目标区域的第一位置作为所述目标对象所在的目标区域的位置，得到所述目标对象的检测结果。
在一些可能的实施方式中，所述基于所述输入图像，确定目标对象的检测结果之后，所述方法还包括以下至少一种：
响应于所述目标对象的检测结果中仅包括指示灯底座对应的检测结果，确定所述指示灯为故障状态；
响应于所述目标对象的检测结果中仅包括点亮状态的指示灯对应的检测结果，确定采集所述输入图像的场景状态为黑暗状态。
在一些可能的实施方式中，所述基于所述目标对象的检测结果，对所述输入图像中所述目标对象所在的目标区域进行识别，得到所述目标对象的指示信息，包括：
基于所述目标对象的检测结果中所述目标对象的类型，确定与所述目标对象匹配的分类器；
利用匹配的分类器，对所述输入图像中所述目标区域的图像特征进行识别，得到所述目标对象的指示信息。
在一些可能的实施方式中，所述基于所述目标对象的检测结果，对所述输入图像中所述目标对象所在的目标区域进行识别，得到所述目标对象的指示信息，包括：
响应于所述目标对象的类型为指示灯底座，确定匹配的分类器包括用于对所述指示灯底座中的指示灯的排列方式进行识别的第一分类器，利用所述第一分类器，对所述目标对象所在的目标区域的图像特征进行识别，确定所述指示灯底座中的指示灯的排列方式；和/或，
确定匹配的分类器包括用于对所述指示灯所在场景进行识别的第二分类器，利用所述第二分类器，对所述目标对象所在的目标区域的图像特征进行识别，确定所述指示灯所在的场景信息。
在一些可能的实施方式中，所述基于所述目标对象的检测结果，对所述输入图像中所述目标对象所在的目标区域进行识别，得到所述目标对象的指示信息，包括：
响应于所述目标对象的类型为圆斑灯或者行人灯，确定匹配的分类器包括用于对圆斑灯的颜色属性进行识别的第三分类器；
利用所述第三分类器，对所述目标对象所在的目标区域的图像特征进行识别，确定所述圆斑灯或者行人灯的颜色属性。
在一些可能的实施方式中，所述基于所述目标对象的检测结果，对所述输入图像中所述目标对象所在的目标区域进行识别，得到所述目标对象的指示信息，包括：
响应于所述目标对象的类型为箭头灯，确定匹配的分类器包括用于对箭头灯的颜色属性的第四分类器以及方向属性进行识别的第五分类器；
利用所述第四分类器和所述第五分类器，对所述目标对象所在的目标区域的图像特征进行识别，分别确定所述箭头灯的颜色属性和方向属性。
在一些可能的实施方式中，所述基于所述目标对象的检测结果，对所述输入图像中所述目标对象所在的目标区域进行识别，得到所述目标对象的指示信息，包括：
响应于所述目标对象的类型为数字灯，确定匹配的分类器包括用于对数字灯的颜色属性的第六分类器以及数值属性进行识别的第七分类器；
基于所述第六分类器和所述第七分类器，对所述目标对象所在的目标区域的图像特征进行识别，分别确定所述数字灯的颜色属性和数值属性。
在一些可能的实施方式中，响应于所述输入图像中包括至少两个指示灯底座，所述方法还包括：
针对第一指示灯底座，确定与所述第一指示灯底座匹配的点亮状态的指示灯；第一指示灯底座为所述至少两个指示灯底座中之一；
将所述第一指示灯底座的指示信息、与所述第一指示灯底座匹配的点亮状态的指示灯的指示信息进行组合，得到组合后的指示信息。
在一些可能的实施方式中，所述确定与所述第一指示灯底座匹配的点亮状态的指示灯，包括：
基于所述目标对象的检测结果中目标对象所在的目标区域的位置，确定所述至少一个点亮状态的指示灯所在的目标区域与所述第一指示灯底座所在的目标区域之间相交的第一面积，以及，确定所述至少一个点亮状态的指示灯所在的目标区域的第二面积；
响应于存在点亮状态的第一指示灯和所述第一指示灯底座之间的所述第一面积，与所述点亮状态的第一指示灯的第二面积之间的比值大于设定面积阈值，确定所述点亮状态的第一指示灯与所述第一指示灯底座相匹配；
其中，所述点亮状态的第一指示灯为所述至少一个点亮状态的指示灯中之一。
根据本公开的第二方面，提供了一种驾驶控制方法，其包括：
利用智能驾驶设备中的图像采集设备采集行驶图像；
对所述行驶图像执行第一方面中所述的指示灯指示信息识别方法，得到针对所述行驶图像的指示信息；
利用所述指示信息，生成所述智能驾驶设备的控制指令。
根据本公开的第三方面，提供了一种指示灯的指示信息识别装置，其包括：
获取模块，其用于获取输入图像；
检测模块，其用于基于所述输入图像，确定目标对象的检测结果，所述目标对象包括指示灯底座、点亮状态的指示灯中的至少一种，所述检测结果包括所述目标对象的类型、所述输入图像中所述目标对象所在的目标区域的位置；
识别模块，其用于基于所述目标对象的检测结果，对所述输入图像中所述目标对象所在的目标区域进行识别，得到所述目标对象的指示信息。
在一些可能的实施方式中，所述确定模块还用于：
提取所述输入图像的图像特征；
基于所述输入图像的图像特征，确定所述目标对象的至少一个候选区域中每个候选区域的第一位置；
基于所述输入图像中每个候选区域对应的第一位置处的图像特征，确定每个候选区域的中间检测结果，所述中间检测结果包括所述目标对象的预测类型和所述目标对象为所述预测类型的预测概率；所述预测类型为指示灯底座、N种点亮状态的指示灯中的任一种，N为正整数；
基于至少一个候选区域中每个候选区域的中间检测结果以及每个候选区域的第一位置，确定所述目标对象的检测结果。
在一些可能的实施方式中，所述确定模块还用于：针对每个候选区域，基于该候选区域对应的第一位置处的图像特征，对该候选区域内的所述目标对象进行分类，得到所述目标对象为所述至少一种预设类型中每种预设类型的预测概率；其中，所述预设类型包括指示灯底座、N种点亮状态的指示灯中的至少一种，N为正整数；
将所述至少一种预设类型中预测概率最高的预设类型，作为该候选区域内所述目标对象的预测类型，并得到所述预测类型的预测概率。
在一些可能的实施方式中，所述确定模块还用于：在基于至少一个候选区域中每个候选区域的中间检测结果以及每个候选区域的第一位置，确定所述目标对象的检测结果之前，基于所述输入图像的图像特征，确定每个候选区域的第一位置的位置偏差；
利用每个候选区域对应的位置偏差，调整每个候选区域的第一位置。
在一些可能的实施方式中，所述确定模块还用于在所述目标对象的候选区域为至少两个的情况下，基于至少两个候选区域中每个候选区域的中间检测结果，或者，基于每个候选区域的中间检测结果以及每个候选区域的第一位置，从所述至少两个候选区域中筛选出目标区域；
将所述目标区域内所述目标对象的预测类型作为所述目标对象的类型，将所述目标区域的第一位置作为所述目标对象所在的目标区域的位置，得到所述目标对象的检测结果。
在一些可能的实施方式中，所述确定模块还用于在所述目标对象的检测结果中仅包括指示灯底座 对应的检测结果的情况下，确定所述指示灯为故障状态；
在所述目标对象的检测结果中仅包括点亮状态的指示灯对应的检测结果的情况下，确定采集所述输入图像的场景状态为黑暗状态。
在一些可能的实施方式中，所述识别模块还用于基于所述目标对象的检测结果中所述目标对象的类型，确定与所述目标对象匹配的分类器；
利用匹配的分类器，对所述输入图像中所述目标区域的图像特征进行识别，得到所述目标对象的指示信息。
在一些可能的实施方式中，所述识别模块还用于在所述目标对象的类型为指示灯底座的情况下，确定匹配的分类器包括用于对所述指示灯底座中的指示灯的排列方式进行识别的第一分类器，利用所述第一分类器，对所述目标对象所在的目标区域的图像特征进行识别，确定所述指示灯底座中的指示灯的排列方式；和/或，
确定匹配的分类器包括用于对所述指示灯所在场景进行识别的第二分类器，利用所述第二分类器，对所述目标对象所在的目标区域的图像特征进行识别，确定所述指示灯所在的场景信息。
在一些可能的实施方式中，所述识别模块还用于在所述目标对象的类型为圆斑灯或者行人灯的情况下，确定匹配的分类器包括用于对圆斑灯的颜色属性进行识别的第三分类器；
利用所述第三分类器，对所述目标对象所在的目标区域的图像特征进行识别，确定所述圆斑灯或者行人灯的颜色属性。
在一些可能的实施方式中，所述识别模块还用于在所述目标对象的类型为箭头灯的情况下，确定匹配的分类器包括用于对箭头灯的颜色属性的第四分类器以及方向属性进行识别的第五分类器；
利用所述第四分类器和所述第五分类器，对所述目标对象所在的目标区域的图像特征进行识别，分别确定所述箭头灯的颜色属性和方向属性。
在一些可能的实施方式中，所述识别模块还用于在所述目标对象的类型为数字灯的情况下，确定匹配的分类器包括用于对数字灯的颜色属性的第六分类器以及数值属性进行识别的第七分类器；
基于所述第六分类器和所述第七分类器，对所述目标对象所在的目标区域的图像特征进行识别，分别确定所述数字灯的颜色属性和数值属性。
在一些可能的实施方式中，所述装置还包括匹配模块，其用于在所述输入图像中包括至少两个指示灯底座的情况下，针对第一指示灯底座，确定与所述第一指示灯底座匹配的点亮状态的指示灯；第一指示灯底座为所述至少两个指示灯底座中之一；
将所述第一指示灯底座的指示信息、与所述第一指示灯底座匹配的点亮状态的指示灯的指示信息进行组合，得到组合后的指示信息。
在一些可能的实施方式中，所述匹配模块还用于：
基于所述目标对象的检测结果中目标对象所在的目标区域的位置，确定所述至少一个点亮状态的指示灯所在的目标区域与所述第一指示灯底座所在的目标区域之间相交的第一面积，以及，确定所述至少一个点亮状态的指示灯所在的目标区域的第二面积；
在存在点亮状态的第一指示灯和所述第一指示灯底座之间的所述第一面积，与所述点亮状态的第一指示灯的第二面积之间的比值大于设定面积阈值的情况下，确定所述点亮状态的第一指示灯与所述第一指示灯底座相匹配；
其中，所述点亮状态的第一指示灯为所述至少一个点亮状态的指示灯中之一。
根据本公开的第四方面，提供了一种驾驶控制装置，其包括：
图像采集模块，其设置在智能驾驶设备中，并用于采集所述智能驾驶设备的行驶图像；
图像处理模块，其用于对所述行驶图像执行如第一方面中任意一项所述的指示灯指示信息识别方法，得到针对所述行驶图像的指示信息；
控制模块，其用于利用所述指示信息，生成所述智能驾驶设备的控制指令。
根据本公开的第五方面，提供了一种电子设备，其包括：
处理器；
用于存储处理器可执行指令的存储器；
其中，所述处理器被配置为调用所述存储器存储的指令，以执行第一方面或者第二方面中任意一项所述的方法。
根据本公开的第六方面，提供了一种计算机可读存储介质，其上存储有计算机程序指令，所述计算机程序指令被处理器执行时实现第一方面或者第二方面中任意一项所述的方法。
根据本公开的第七方面，提供了一种计算机程序，包括计算机可读代码，当所述计算机可读代码在电子设备中运行时，所述电子设备中的处理器执行用于实现第一方面或者第二方面中任意一项所述的方法。
本公开实施例可以首先对输入图像进行目标检测处理，得到目标对象检测结果，其中目标对象的检测结果可以包括目标对象的位置以及类型等信息，再进一步根据目标对象的检测结果执行目标对象的指示信息的识别。本公开通过将目标对象的检测过程，划分为对指示灯底座和点亮状态的指示灯这两个检测过程，在检测过程中实现了对目标对象的首次区分，后续基于目标对象的检测结果进行进一步识别时，有利于降低在识别目标对象的指示信息的过程中的识别复杂度，降低识别难度，可以简单方便的实现不同情况下的对各类型的指示灯的检测识别。
应当理解的是，以上的一般描述和后文的细节描述仅是示例性和解释性的，而非限制本公开。
根据下面参考附图对示例性实施例的详细说明，本公开的其它特征及方面将变得清楚。
附图说明
此处的附图被并入说明书中并构成本说明书的一部分，这些附图示出了符合本公开的实施例，并与说明书一起用于说明本公开的技术方案。
图1示出根据本公开实施例的一种指示灯的指示信息识别方法的流程图；
图2(a)示出交通灯的不同显示状态；
图2(b)示出交通灯底座的不同排列方式；
图2(c)示出交通灯的不同应用场景；
图2(d)示出多种交通灯的类型；
图2(e)示出不同情况的组合交通灯的示意图；
图3示出根据本公开实施例的一种指示灯的指示信息识别方法中步骤S20的流程图；
图4示出根据本公开实施例通过区域候选网络执行目标检测的示意图；
图5示出根据本公开实施例的一种指示灯的指示信息识别方法中步骤S30的流程图；
图6示出本公开实施例不同的目标对象的分类检测示意图；
图7示出多个底座的交通灯结构示意图；
图8示出根据本公开实施例的一种指示灯的指示信息识别方法的另一流程图；
图9示出根据本公开实施例的一种驾驶控制方法的流程图；
图10示出根据本公开实施例的一种指示灯的指示信息识别装置的框图；
图11示出根据本公开实施例的一种驾驶控制装置的框图；
图12示出根据本公开实施例的一种电子设备的框图；
图13示出根据本公开实施例的一种电子设备的另一框图。
具体实施方式
以下将参考附图详细说明本公开的各种示例性实施例、特征和方面。附图中相同的附图标记表示功能相同或相似的元件。尽管在附图中示出了实施例的各种方面，但是除非特别指出，不必按比例绘制附图。
在这里专用的词“示例性”意为“用作例子、实施例或说明性”。这里作为“示例性”所说明的任何实施例不必解释为优于或好于其它实施例。
本文中术语“和/或”，仅仅是一种描述关联对象的关联关系，表示可以存在三种关系，例如，A和/或B，可以表示：单独存在A，同时存在A和B，单独存在B这三种情况。另外，本文中术语“至少一种”表示多种中的任意一种或多种中的至少两种的任意组合，例如，包括A、B、C中的至少一种， 可以表示包括从A、B和C构成的集合中选择的任意一个或多个元素。
另外，为了更好地说明本公开，在下文的具体实施方式中给出了众多的具体细节。本领域技术人员应当理解，没有某些具体细节，本公开同样可以实施。在一些实例中，对于本领域技术人员熟知的方法、手段、元件和电路未作详细描述，以便于凸显本公开的主旨。
本公开实施例提供的指示灯的指示信息识别方法可以用于执行不同类型的指示灯的指示信息检测，其中，该指示灯的指示信息识别方法可以由任意具有图像处理功能的电子设备执行，例如由终端设备或服务器或其它处理设备执行，其中，终端设备可以为用户设备(User Equipment，UE)、移动设备、用户终端、终端、蜂窝电话、无绳电话、个人数字处理(Personal Digital Assistant，PDA)、手持设备、计算设备、车载设备、可穿戴设备等。或者，在一些可能的实施方式中，指示灯的指示信息识别方法还可以应用在智能驾驶设备中，如智能飞行设备、智能车辆、导盲设备中，用于智能驾驶设备的智能控制，另外，在一些可能的实现方式中，该指示灯的指示信息识别方法可以通过处理器调用存储器中存储的计算机可读指令的方式来实现。本公开实施例提供的指示灯的指示信息识别方法可以应用于指示灯的指示信息识别、检测等场景，例如，可以用于自动驾驶、监控等应用场景中指示灯的指示信息识别，本公开不对具体的应用场景进行限制。
图1示出根据本公开实施例的指示灯的指示信息识别方法的流程图，如图1所示，指示灯的指示信息识别方法包括：
S10：获取输入图像。
在一些可能的实施方式中，输入图像可以为关于指示灯的图像，指示灯可以包括交通指示灯(例如红绿灯)、应急指示灯(例如闪烁状态的指示灯)、方向指示灯中的至少一种，在其他实施例中，也可以为其他类型的指示灯。
本公开可以实现输入图像中的指示灯的指示信息的识别。其中，输入图像可以是通过图像采集设备采集的图像，例如可以为通过设置在车辆中的图像采集设备采集的路面行驶图像，或者也可以为通过布设的摄像头采集的图像，或者在其他实施例中也可以通过手持终端设备或者其他设备采集的图像，或者，输入图像也可以是从获取的视频流中选择出的图像帧，本公开对此不作具体限定。
S20：基于输入图像，确定目标对象的检测结果，目标对象包括指示灯底座、点亮状态的指示灯中的至少一种，检测结果包括目标对象的类型、输入图像中目标对象所在的目标区域的位置。
在一些可能的实施方式中，在得到输入图像的情况下，可以对输入图像中的目标对象进行检测识别，得到关于目标对象的检测结果。其中检测结果可以包括目标对象的类型以及位置信息。本公开实施例，可以通过神经网络实现输入图像中目标对象的目标检测，得到检测结果。通过该神经网络可以实现输入图像中指示灯底座类型、点亮状态的指示灯类型，以及底座位置、点亮指示灯位置中至少一种信息的检测。其中可以通过任意可以实现目标对象的检测以及可以实现分类的神经网络得到输入图像的检测结果。其中神经网络可以为卷积神经网络。
实际应用中，采集的输入图像中包括的指示灯的形态可以为多种，以交通指示灯(以下简称交通灯)为例，交通灯的形态可以是各种各样的，在交通灯的类型为圆斑灯的情况下，如图2(a)-图2(e)中分别示出交通灯的多种显示状态的示意图，其中图2(a)中显示交通灯不同的显示状态，交通灯底座的形状，本公开对此不作限定。
由于实际生活中，指示灯底座中可能包括多种颜色状态的指示灯，因此对应的指示灯显示状态也会有多种。以图2(a)中的交通灯为例进行说明，其中以第一组交通灯为例进行说明，其中L表示交通灯，D表示交通灯底座，从图2(a)可以看出，第一组为交通灯中的红黄绿三个灯均为灭灯状态，此时可能为故障状态。第二组交通灯为红灯被点亮的状态、第三个交通灯为黄灯被点亮的状态，以及第四组交通灯为绿灯被点亮的状态。在识别目标对象过程中，可以对是否为点亮状态的指示灯以及点亮状态的指示灯的颜色进行识别。其中，红、黄、绿的文字仅为示意性的表示相应的颜色的交通灯处于点亮状态。
图2(b)示出交通灯底座的不同排列方式，一般地，交通灯或者其他类型的指示灯都可以被装设在指示灯底座上，如图2(b)所示，其中交通灯在底座上的排列方式可以包括横向排列、纵向排列或 者单个灯。因此，在识别目标对象过程中，也可以对指示灯的排列方式进行识别，上述仅为交通灯在底座上排列方式的示例性说明，在其他实施例中，指示灯在底座上的排列方式也可以包括其他类型的排列方式。
图2(c)示出交通灯的不同应用场景，在实际应用中交通灯等指示灯可以设置在道路的十字路口、高速路口、急转路口、安全警示位置、或者行进通道上，因此，对于指示灯的识别还可以对指示灯的应用场景进行判断识别，如图2(c)中现实的应用场景依次为标注有“电子收费系统(Electronic Toll Collection，ETC)”标识的高速路口、标注有“警告信号”等警告标识的急转路口或者其他危险场景以及一般场景。上述场景为示例性说明，本公开对此不作具体限定。
图2(d)示出多种交通灯的类型，一般地，根据需求或者场景的需要，交通灯或者其他指示灯的形状也各不相同，例如图2(d)中依次示出的包含有箭头形状的箭头灯、包含有圆斑形状的圆斑灯、包含有行人标识的行人灯、或者包含有数字数值的数字灯，同时各类型的灯还可以具有不同的颜色，本公开对此并不限定。
图2(e)示出不同情况的组合交通灯的示意图。其中，例如不同箭头方向的箭头灯的组合，数字灯和行人灯的组合，其中还具有颜色等指示信息。如上所述，实际应用中，存在各式各样的指示灯，本公开可以实现对于各式各样的指示灯的指示信息的识别。
正是因为上述情况的复杂性，本公开实施例首先可以通过对输入图像进行目标对象的检测，从而确定输入图像中目标对象的检测结果，在基于该检测结果进一步得到目标对象的指示信息。例如，通过对输入图像执行目标检测，可以检测出输入图像中的目标对象的类型以及位置，或者，检测结果中也可以包括目标对象的类型的概率。在得到上述检测结果的情况下，再进一步根据检测到的目标对象的类型执行分类检测，得到目标对象的指示信息，如亮灯颜色、数值、方向、场景等信息。
本公开实施例可以将检测目标(即目标对象)的类型分为指示灯底座和点亮状态的指示灯两个部分，其中点亮状态的指示灯可以包括N种类型，例如指示灯的类型可以包括上述数字灯、行人灯、箭头灯和圆斑灯中的至少一种。因此在执行目标对象的检测时，可以确定输入图像中包括的每个目标对象为N+1种类型(底座以及N种点亮指示灯)中的任意一种类型。或者在其他实施方式中也可以包括其他类型的指示灯，本公开对此不作具体限定。
示例性的，本公开中对于灭灯状态的指示灯可以不执行检测，在未检测到指示灯底座以及点亮状态的指示灯的情况下，可以认为输入图像内不存在指示灯，故可以无需执行S30中进一步识别目标对象的指示信息的过程。另外，在检测出指示灯底座、但未检测出点亮状态的指示灯的情况下，也可以视为存在灭灯状态的指示灯，这种情况也无需进行目标对象的指示信息的识别。
S30：基于目标对象的检测结果，对输入图像中目标对象所在的目标区域进行识别，得到目标对象的指示信息。
在一些可能的实施方式中，在得到目标对象的检测结果的情况下，可以进一步检测目标对象的指示信息，其中，指示信息用于描述目标对象的相关属性。在智能驾驶领域中，目标对象的指示信息可以用于指示智能驾驶设备基于指示信息生成控制指令。例如对于类型为底座的目标对象，可以识别出指示灯的排列方式以及应用场景中的至少一种，对于类型为点亮状态的指示灯的目标对象，可以识别出指示灯的亮灯颜色、箭头的指示方向、数字的数值等信息中的至少一种。
基于本公开的实施例，可以通过首先对底座和点亮状态的指示灯进行检测，并基于得到的检测结果进一步分类识别目标对象的指示信息，即可以不直接通过分类器对目标对象的类型、位置、以及各类指示信息等信息一起进行分类识别，而是根据目标对象的类型等检测结果执行指示信息的分类识别，有利于降低在识别目标对象的指示信息的过程中的识别复杂度，降低识别难度，同时还可以简单方便的实现不同情况下的对各类型的指示灯的检测识别。
下面结合附图对本公开实施例的具体过程进行分别说明。图3示出根据本公开实施例的一种指示灯的指示信息识别方法中步骤S20的流程图。其中，基于输入图像，确定目标对象的检测结果(步骤S20)，可以包括：
S21：提取输入图像的图像特征；
在一些可能的实施方式中，在得到输入图像的情况下，可以对输入图像执行特征提取处理，得到输入图像的图像特征。其中，可以通过特征提取算法得到输入图像中的图像特征，也可以通过经过训练可以实现特征提取的神经网络提取图像特征。例如本公开实施例可以采用卷积神经网络得到输入图像的图像特征，可以通过对输入图像执行至少一层卷积处理得到对应的图像特征，其中卷积神经网络可以包括视觉几何组(Visual Geometry Group，VGG)网络、残差网络、金字塔特征网络中的至少一种，但不作为本公开的具体限定，也可以通过其他方式得到图像特征。
S22：基于输入图像的图像特征，确定目标对象的至少一个候选区域中每个候选区域的第一位置；
在一些可能的实施方式中，可以根据输入图像的图像特征，检测输入图像中目标对象所在的位置区域，即得到每个目标对象的候选区域的第一位置。其中，针对每个目标对象，可以得到至少一个候选区域，对应的可以得到每个候选区域的第一位置，本公开实施例的第一位置可以利用候选区域的对角顶点位置的坐标来表示，但本公开对此不作具体限定。
图4示出根据本公开实施例执行目标检测的示意图。其中，执行目标检测所采用的目标检测网络可以包括基础网络(base network)模块、区域候选网络(Region Proposal Network，RPN)模块、分类模块，其中，基础网络模块用于执行输入图像(image)的特征提取处理，得到输入图像的图像特征。区域候选网络模块用于基于输入图像的图像特征，检测与输入图像中目标对象的候选区域(Region of Interest，ROI)，分类模块用于基于候选区域的图像特征对候选区域内的目标对象的类型进行判断，得到输入图像中目标区域(Box)中目标对象的检测结果。示例性的，目标对象的检测结果包括目标对象的类型以及目标区域的位置，目标对象的类型例如为底座、点亮状态的指示灯(如圆斑灯、箭头灯、行人灯、数字灯)、背景(background)中的任一种。其中，背景可以理解为输入图像中除底座、点亮状态的指示灯所在区域之外的图像区域。
在一些可能的实施方式中，区域候选网络可以针对输入图像中的每个目标对象得到至少一个ROI，通过后续的后处理可以从中选择出精确度最高的ROI。
S23：基于输入图像中每个候选区域对应的第一位置处的图像特征，确定每个候选区域的中间检测结果，中间检测结果包括目标对象的预测类型和目标对象为预测类型的预测概率；预测类型为指示灯底座、N种点亮状态的指示灯中的任一种，N为正整数。
在得到每个目标对象的至少一个候选区域(如第一候选区域或者第二候选区域)的情况下，可以进一步分类识别候选区域内的目标对象的类型信息，即可以得到候选区域内的目标对象的预测类型以及针对该预测类型的预测概率。其中预测类型可以为上述N+1种类型中的一种，例如可以为底座、圆斑灯、箭头灯、行人灯、数字灯中任一种。也就是说，可以预测候选区域内的目标对象的类型是底座还是N种点亮状态的指示灯中的一种。
其中，步骤S23可以包括：针对每个候选区域，基于该候选区域对应的第一位置处的图像特征，对该候选区域内的目标对象进行分类，得到目标对象为至少一种预设类型中每种预设类型的预测概率；其中，预设类型包括指示灯底座、N种点亮状态的指示灯中的至少一种，N为正整数；将至少一种预设类型中预测概率最高的预设类型，作为该候选区域内目标对象的预测类型，并得到预测类型的预测概率。
在一些可能的实施方式中，在得到每个目标对象的至少一个候选区域的情况下，可以根据候选区域的第一位置得到输入图像的图像特征中与该第一位置对应的图像特征，并将得到的该图像特征确定为候选区域的图像特征。进一步地，可以根据每个候选区域的图像特征预测该候选区域内的目标对象为各预设类型的预测概率。
其中，针对每个候选区域，可以对候选区域内的图像特征执行分类识别，对应的可以得到每个候选区域针对每种预设类型的预测概率，其中预设类型为上述N+1种类型，如底座以及N种指示灯类型。或者，在其他实施例中预设类型也可以为N+2种类型，相对于N+1种类型的情况还进一步包括背景类型，但本公开对此不作具体限定。
在得到候选区域内的目标对象为各预设类型的预测概率的情况下，可以将预测概率最高的预设类型确定为该候选区域内的目标对象的预测类型，对应的该最高的预测概率为相应的预测类型的预测概 率。
在一些可能的实施方式中，在对候选区域的目标对象执行类型分类检测之前，可以对每个候选区域的图像特征进行池化处理，从而使得每个候选区域的图像特征的尺度相同，例如对于每一个ROI，可以将图像特征的尺寸缩放到7*7，但本公开对此不作具体限定。在池化处理后，可以对池化处理后的图像特征进行分类处理，得到针对每个目标对象的每个候选框对应的中间检测结果。
在一些可能的实施方式中，步骤S23中对每个候选区域的图像特征进行分类处理可以利用一个分类器实现，也可以利用多个分类器实现。例如根据一个分类器得到候选区域针对每个预设类型的预测概率，或者也可以利用N+1或者N+2个分类器，分别检测候选区域针对各个类型的预测概率，该N+1或者N+2个分类器与预设类型一一对应，即每个分类器可以用于得到相应的预设类型的预设结果。
在一些可能的实施方式中，在对候选区域执行分类处理时，还可以利用卷积层对候选区域的图像特征(或者池化后的图像特征)输入至第一卷积层中执行卷积处理，得到维度为a×b×c的第一特征图，b和c分别表示第一特征图的长度和宽度，a表示第一特征图的通道数，并且a的数值为预设类型的总数(如N+1)，而后对第一特征图执行全局池化处理得到与第一特征图对应的第二特征图，该第二特征图的维度为a×d，将该第二特征图输入至softmax函数中，同样可以得到a×d维的第三特征图，其中d为大于或者等于1的整数。在一个示例中，d表示第三特征图的列数，如可以为1，对应的得到的第三特征图中的元素表示候选区域内的目标对象为各个预设类型的预测概率，每个元素对应的数值可以为该预测概率的概率值，该概率值的顺序与设定的预设类型的顺序对应，或者第三特征图中的每个元素可以由预设类型的标识以及对应的预测概率构成，从而方便的确定预设类型和预测概率的对应关系。
在另一个示例中，d也可以为大于1的其他整数值，可以根据第三特征图中的第一预设列数的元素得到预设类型对应的预测概率。该第一预设列数可以预先设定的值，如可以为1，但不作为本公开的具体限定。
通过上述配置，可以得到每个目标对象的每个候选区域的中间检测结果，进一步的可以利用中间检测结果得到每个目标对象检测结果。
S24：基于至少一个候选区域中每个候选区域的中间检测结果以及每个候选区域的第一位置，确定目标对象的检测结果。
如上述实施例所述，可以得到针对每个目标对象的所有候选区域对应的中间检测结果(如候选区域的第一位置、候选区域内的目标对象的预测类型和预测概率)，进一步地，可以根据目标对象的每个候选区域的中间检测结果确定目标对象的最终检测结果，即目标对象的候选区域的位置、类型等信息。
在此需要说明的是，本公开实施例可以将每个目标对象的候选区域的第一位置作为该候选区域的位置，或者也可以对第一位置进行优化，得到更精确的第一位置。其中，本公开实施例还可以通过每个候选区域的图像特征得到相应的候选区域的位置偏差，并根据该位置偏差调整候选区域的第一位置。其中，可以将每个目标对象的候选区域的图像特征输入至第二卷积层，得到第四特征图，第四特征图的维度为e×b×c，其中，b和c分别表示第四特征图和第三特征图的长度和宽度，同时b和c也可以为候选区域的图像特征的长度和宽度，e表示第四特征图的通道数，e可以为大于或者等于1的整数，例如e可以取值为4。进一步地，通过对第四特征图执行全局池化处理，可以得到第五特征图，该第五特征图可以为长度为e的特征向量，如e等于4，此时第五特征图中元素即为相应的候选区域对应的位置偏差。或者，在其他实施例中，第五特征图的维度可以为e×f，f为大于或者等于1的数值，表示第五特征图的列数。此时可以根据第三特征图中的预设位置区域中的元素得到候选区域的位置偏差。其中预设位置区域可以为预先设定的位置区域，如第1-4行第1列的元素，但不作为本公开的具体限定。
示例性的，候选区域的第一位置可以表示为两个对角的顶点位置的横纵坐标值，第五特征图中的元素则可以为该两个顶点的横纵坐标值的位置偏移量。在得到第五特征图之后，即可以将候选区域的第一位置按照该第五特征图中的相应位置偏差进行调整，得到精确度更高的第一位置。其中， 第一卷积层和第二卷积层为两个不同的卷积层。
由于在执行目标对象的检测过程中，针对输入图像中的每个目标对象可以检测到至少一个候选区域，本公开实施例可以从该至少一个候选区域中筛选出目标对象的目标区域。
在针对输入图像的任一目标对象仅检测出一个候选区域的情况下，可以判断基于该候选区域确定的目标对象的预测类型的预测概率是否大于概率阈值，如果大于概率阈值，则可以将该候选区域确定为该目标对象的目标区域，并且将该候选区域对应得到的预测类型确定为目标对象的类型。如果基于该候选区域确定的目标对象的预测类型的预测概率小于概率阈值，则舍弃该候选区域，确定该候选区域内的对象并不存在所要检测的任一目标对象。
或者，在针对输入图像的一个或多个目标对象检测出多个候选区域的情况下，可以基于每个候选区域的中间检测结果，或者基于每个候选区域的中间检测结果每个候选区域的第一位置，从该多个候选区域中筛选出目标区域，并将目标区域内目标对象的预测类型作为目标对象的类型，将目标区域的第一位置作为目标对象所在的目标区域的位置，得到目标对象的检测结果。
示例性的，基于候选区域的中间检测结果筛选出目标区域的步骤可以包括：从目标对象的多个候选区域中选择出对应的预测概率最高的候选区域，并在该最高的预测概率大于概率阈值的情况下，将最高的预测概率对应的候选区域的第一位置(或者调整后的第一位置)作为该目标对象的目标区域，以及该最高的预测概率对应的预测类型确定为目标对象的类型。
示例性的，基于候选区域的第一位置筛选出目标对象的目标区域的步骤可以包括：利用非极大值抑制算法(Non-maximum suppression，NMS)从多个候选区域中选择出目标对象的目标区域。其中，可以得到输入图像中目标对象的多个候选区域中选择出预测概率最大的候选区域，以下称为第一候选区域。并根据第一候选区域的第一位置以及其余候选区域的第一位置，确定其余候选区域分别与第一候选区域之间的重叠区域值(Intersection over Union，IOU)，如果其余候选区域中任一候选区域与第一候选区域之间的IOU大于面积阈值，则舍弃该任一候选区域。如果比较IOU之后，其余的候选区域均被舍弃，则第一候选区域则为目标对象的目标区域，同时基于第一候选区域得到的目标对象的预测类型可以为该目标对象的类型。如果其余的候选区域中存在至少一个第二候选区域与第一候选区域之间的IOU值小于面积阈值，则可以将第二候选区域中预测概率最高的候选区域重新作为新的第一候选区域，继续得到第二候选区域中的其余候选区域与该新的第一候选区域的IOU，同样舍弃IOU大于面积阈值的第二候选区域，直至不存在与第一候选区域(或者新的候选区域)的IOU大于面积阈值的候选区域。通过上述方式得到的各第一候选区域可以确定为各个目标对象的目标区域。
或者，在其他可能的实施方式中，也可以通过概率阈值从每个目标对象的候选区域中筛选出预测概率大于概率阈值的候选区域，而后在通过上述NMS算法得到每个目标对象的目标区域，同时得到目标区域内针对目标对象的预测类型，即确定目标对象的检测结果。
在此需要说明的是，上述通过第一位置执行检测结果的确定过程，也可以通过调整后的第一位置来执行目标对象的检测结果的确定，具体原理相同，在此不做重复说明。
基于上述实施例，可以得到输入图像中存在的目标对象的检测结果，即可以方便的确定出目标对象的类型以及相应的位置。其中，通过上述目标检测可以得到针对每个目标对象(例如点亮状态的指示灯、指示灯底座)的检测框(候选区域)，例如对于点亮状态的指示灯，检测结果中可以包括输入图像中的点亮状态的指示灯的位置以及该指示灯的类型，例如检测结果可以表示为(x1,y1,x2,y2,label1,score1)。其中(x1,y1)，(x2,y2)是点亮状态的指示灯的目标区域的位置坐标(两个对角的点的坐标)，label1表示点亮状态的指示灯的类型标识(1到N+1中的一个，如2，可以表示为数字灯)，score1表示该检测结果的置信度(即，预测概率)。
对于指示灯底座，检测结果表示为(x3,y3,x4,y4,label2,scor12)。其中(x3,y3)，(x4,y4)是底座的目标区域的位置坐标(两个对角的点的坐标)，label2表示底座的类型标识(1到N中的一个，如1)，score2表示该检测结果的置信度。其中底座的标识可以为1，其余N个标识可以为点亮状态的指示灯的N种类型，在一些可能的实施方式中，还可以标识N+2，用以表示背景的目标区域，本公开对此不作具体限定
基于上述，即可以简单方便的得到针对目标对象的检测结果。同时由于检测结果中已经包括了指示灯或者底座的类型信息，可以减少后续分类器的分类压力。
在一些可能的实施方式中，在得到输入图像中目标对象的检测结果的情况下，还可以基于该检测结果进一步确定指示灯是否出现故障，或者采集输入图像的采集环境等信息。其中，在输入图像的目标对象的结果中，检出的目标对象的类型仅包括指示灯底座，不包括任一类型的点亮状态的指示灯，此时可以确定指示灯为故障状态。例如，在交通信号灯中，如果检测不到任一交通灯处于被点亮的状态，则此时可以确定该交通灯为故障灯，此时可以基于输入图像相关的采集时间、采集地点等信息，执行故障报警操作。例如，向服务器或者其他管理设备发送故障信息，该故障信息可以包括指示灯不亮的故障情况，以及故障灯的位置信息(基于上述采集地点确定)。
或者，在一些实施例中，如果针对输入图像检测到的目标对象的检测结果中，仅包括点亮状态的指示灯，不包括与该点亮状态的指示灯对应的底座，此时可以确定采集输入图像的采集环境为黑暗环境，或者黑暗状态，其中黑暗状态或者黑暗环境是指光亮度小于预设亮度的环境，预设亮度可以根据不同的地点或者不同的天气情况设定，本公开对此不作具体限定。
图5示出根据本公开实施例的一种指示灯的指示信息识别方法中步骤S30的流程图，其中，基于目标对象的检测结果，对输入图像中目标对象所在的目标区域进行识别，得到目标对象的指示信息(步骤S30)，可以包括：
S31：基于目标对象的检测结果中目标对象的类型，确定与目标对象匹配的分类器；
S32：利用匹配的分类器，对输入图像中目标区域的图像特征进行识别，得到目标对象的指示信息。
示例性的，与目标对象匹配的分类器包括至少一种，每种分类器可以对应一种或多种类型的目标对象。
在一些可能的实施方式中，在得到输入图像中的目标对象的检测结果后，可以执行指示信息的分类检测，例如底座中场景信息、指示灯排列方式、指示灯的颜色、描述、指示方向中的至少一种信息的分类识别。其中，本公开实施例可以采用不同的分类器执行不同的指示信息的分类识别，因此可以首先确定执行分类识别的分类器。
图6示出本公开实施例不同的目标对象的分类检测示意图。
在一些可能的实施方式中，在识别出的目标对象的类型为指示灯底座的情况下，可以对底座类型的目标对象进一步执行指示信息的分类识别，得到指示灯的排列方式以及指示灯所在场景中的至少一种指示信息。其中，排列方式可以包括横向排列、纵向排列以及单个指示灯的排列方式等。场景可以包括高速路口、急转路口、一般场景等，上述仅为排列方式和场景的示例性说明，还可以包括其他排列方式或者场景，本公开对此不作具体限定。
在一些可能的实施方式中，在识别出的目标对象的类型为点亮状态的圆斑灯的情况下，可以对该圆斑灯的亮灯颜色进行分类识别，得到亮灯颜色(如红色、绿色、黄色)的指示信息。在识别出的目标对象的类型为点亮状态的数字指示灯的情况下，可以对数值(如1、2、3等)以及亮灯颜色进行分类识别，得到亮灯颜色以及数值的指示信息。在识别出的目标对象的类型的点亮状态的箭头指示灯的情况下，可以对箭头的指示方向(如向前、向左、向右等)以及亮灯颜色进行分类识别，得到亮灯颜色以及指示方向的指示信息。在识别出的目标对象的类型为行人标识的指示灯(行人灯)的情况下，可以对亮灯颜色进行识别，得到亮灯颜色的指示信息。
也就是说，本公开实施例对于目标对象的检测结果中不同类型的目标对象，可以执行不同指示信息的识别，从而更方便且更精确的得到指示灯的指示信息。其中，在执行指示信息的识别时，可以将相应类型的目标对象所在的目标区域对应的图像特征输入至匹配的分类器内，得到分类结果，即得到相应的指示信息。
示例性地，在针对输入图像中的目标对象的检测结果，得到出至少一个目标对象的类型为底座的情况下，确定的匹配分类器包括第一分类器和第二分类器中的至少一种，其中第一分类器用于对底座中指示灯的排列方式进行分类识别，以及第二分类器用于对指示灯所在场景进行分类识别。将该底座 类型的目标对象的目标区域对应的图像特征输入至第一分类器中，可以得到底座中指示灯的排列方式，将该底座类型的目标对象的目标区域对应的图像特征输入至第二分类器中，可以得到指示灯的场景，例如可以采用文本识别的方式得到该场景信息。
在一些可能的实施方式中，在识别出的目标对象的类型为点亮状态的圆斑灯或者行人灯的情况下，确定匹配的分类器包括用于对圆斑灯或者行人灯的颜色属性进行识别的第三分类器。此时可以将圆斑灯类型或者行人灯类型的目标对象对应的目标区域的图像特征输入至匹配的第三分类器中，得到指示灯的颜色属性。
在一些可能的实施方式中，在识别出的目标对象的类型为点亮状态的箭头灯情况下，确定匹配的分类器包括用于对箭头灯的颜色属性的第四分类器以及方向属性进行识别的第五分类器。此时可以将箭头灯类型的目标对象对应的目标区域的图像特征输入至匹配的第四分类器和第五分类器中，利用第四分类器和第五分类器，对目标对象所在的目标区域的图像特征进行识别，分别得到箭头灯的颜色属性和箭头灯的方向属性。
在一些可能的实施方式中，在识别出的目标对象的类型为点亮状态的数字灯情况下，确定匹配的分类器包括用于对数字灯的颜色属性的第六分类器以及对数字灯的数值属性进行识别的第七分类器。此时可以将数字灯类型的目标对象对应的目标区域的图像特征输入至匹配的第六分类器和第七分类器中，基于第六分类器和第七分类器，对目标对象所在的目标区域的图像特征进行识别，分别得到数字灯的颜色属性和数值方向属性。
在此需要说明的是，上述执行颜色属性的分类识别的第三分类器、第四分类器以及第六分类器可以为同一分类器，也可以为不同的分类器，本公开对此不作具体限定。
另外，在一些可能的实施方式中，上述目标区域的图像特征的获取方式可以包括根据对输入图像进行特征提取得到的输入图像的图像特征，以及目标区域的位置信息确定目标区域的图像特征。也就是说，可以直接从输入图像的图像特征中得到与目标区域的位置信息对应的特征，作为目标区域的图像特征。或者，也可以获取输入图像中与目标区域对应的子图像，而后对该子图像执行特征提取，如卷积处理，得到子图像的图像特征，从而确定目标区域的图像特征。上述仅为示例性说明，在其他实施例中也可以通过其他方式得到目标区域的图像特征，本公开对此不作具体限定。
通过上述实施例，可以得到每个目标区域内的目标对象的指示信息。其中可以通过不同的分类器执行不同指示信息的检测，使得分类结果更精确。同时，在得到目标对象的类型的基础上，进一步采用匹配的分类器进行分类识别，而不是采用全部的分类器进行识别，可以有效的利用分类器资源，加快分类速度。
在一些可能的实施方式中，输入图像中可能包括多个指示灯底座，多个点亮状态的指示灯，图7示出多个底座的交通灯结构示意图。在得到的检测结果包括多个指示灯底座以及多个点亮状态的指示灯的情况下，此时，可以对底座和点亮状态的指示灯进行匹配。例如，在图7中包括两个指示灯底座D1和D2，同时每个指示灯底座中可以包括相应的指示灯，在执行指示信息的识别过程中可以确定出包括三个点亮的指示灯，即L1、L2和L3，通过对指示灯底座和点亮状态的指示灯进行匹配，可以确定点亮状态的指示灯L1与指示灯底座D1匹配，同时指示灯L2和L3与底座D2匹配。
图8示出根据本公开实施例的一种指示灯的指示信息识别方法的另一流程图，其中，指示灯的指示信息识别方法还包括指示灯底座与点亮状态的指示灯的匹配过程，具体为：
S41：针对第一指示灯底座，确定与第一指示灯底座匹配的点亮状态的指示灯；第一指示灯底座为至少两个指示灯底座中之一；
其中，在得到的目标对象的检测结果中可以包括针对底座类型的目标对象的目标区域的第一位置以及点亮状态的指示灯所在目标区域的第二位置，本公开实施例可以根据各底座的第一位置和各指示灯的第二位置确定底座和点亮状态的指示灯是否匹配。
其中，可以基于目标对象的检测结果中目标对象所在的目标区域的位置，确定至少一个点亮状态的指示灯所在的目标区域与第一指示灯底座所在的目标区域之间相交的第一面积，以及，确定至少一个点亮状态的指示灯所在的目标区域的第二面积；响应于存在点亮状态的第一指示灯对应的第一面 积，与点亮状态的第一指示灯的第二面积之间的比值大于设定面积阈值，确定点亮状态的第一指示灯与第一指示灯底座相匹配；其中，点亮状态的第一指示灯为至少一个点亮状态的指示灯中之一。
也就是说，针对每个第一指示灯底座，可以根据第一指示灯底座的目标区域的第一位置和每个点亮状态的指示灯的目标区域的第二位置，确定每个底座和每个指示灯的目标区域之间相交或者相重叠的第一面积S1，如果存在一点亮状态的指示灯(第一指示灯)和指示灯底座之间的第一面积S1与点亮状态的指示灯的目标区域的第二面积S2之间的比值(S1/S2)大于面积阈值，则可以确定该第一指示灯与第一指示灯底座匹配。如果能确定出与第一指示灯底座匹配的多个第一指示灯，可以将该多个第一指示灯同时作为第一指示灯底座匹配的指示灯，或者也可以将比值最大的第一指示灯确定为与第一指示灯底座匹配的点亮状态的指示灯。或者也可以将与第一指示灯底座的之间的上述S1/S2比值最大的预设个数个指示灯确定为与第一指示灯底座匹配的指示灯。预设个数可以为2，但不作为本公开的具体限定。另外，面积阈值可以为预先设定的值，如0.8，但不作为本公开的具体限定。
S42：将第一指示灯底座的指示信息、与第一指示灯底座匹配的点亮状态的指示灯的指示信息进行组合，得到组合后的指示信息。
在得到与指示灯底座匹配的点亮状态的指示灯之后，可以将指示灯底座和匹配的点亮状态的指示灯分别得到的指示信息进行组合，得到指示灯的指示信息。如图7所示，可以将指示灯底座D1和点亮状态的指示灯L1的指示信息组合，确定的指示信息包括场景为一般场景，指示灯的排列方式为横向排列，点亮状态的指示灯为圆斑灯，并且颜色为红色。同时还可以将指示灯底座D2和点亮状态的指示灯L2和L3的指示信息组合，确定的指示信息包括场景为一般场景，指示灯的排列方式为横向排列，点亮状态的指示灯为箭头灯，并且包括向右的箭头灯以及向前的箭头灯，其中向右的箭头灯的颜色为红色，向前的箭头灯的颜色为绿色。
另外，对于找不到对应匹配的点亮状态的指示灯的指示灯底座，可以将底座确定为灭灯状态。即可以确定为该底座对应的指示灯为故障灯。对于找不到匹配的指示灯底座的点亮状态的指示灯，单独输出该点亮状态的指示灯对应的指示信息。这种情况，往往是底座的视觉特征不明显造成的，例如夜晚的时候很难检测底座的情况。
另外，在智能驾驶领域，得到的输入图像可以为实时采集的车辆前方或者后方的图像，在得到输入图像中的指示灯对应的指示信息的情况下，还可以基于该得到的指示信息进一步生成驾驶设备的驾驶参数的控制指令，该驾驶参数可以包括驾驶速度、驾驶方向、控制模式、停止等驾驶状态。
为了更清楚的体现本公开实施例，下面举例说明本公开实施例获取指示信息的过程。本公开实施例所采用的算法模型可以包括两部分，一部分为参照图4所示的执行目标检测的目标检测网络，另一部分为执行指示信息的分类识别的分类网络。其中，结合图4所示，目标检测网络可以包括基础网络(base network)模块、区域候选网络(RPN)模块、分类模块，其中，基础网络模块用于执行输入图像的特征提取处理，得到输入图像的图像特征。区域候选网络模块用于基于输入图像的图像特征，检测与输入图像中目标对象的候选区域(ROI)，分类模块用于基于候选区域的图像特征对候选区域内的目标对象的类型进行判断，得到输入图像的目标对象的检测结果。
其中，目标检测网络的输入为输入图像，输出为若干目标对象的2D检测框(即目标对象的目标区域)，每个检测框可以表示为是(x1,y1,x2,y2,label,score)。其中x1,y1,x2,y2是检测框的位置坐标，label是类别(取值范围是1到N+1，第一个类别代表底座，其他类别代表各种点亮状态的指示灯。
目标检测的过程可以包括：将输入图像输入至Base Network，得到输入图像的图像特征。使用区域候选网络(Region Proposal Network，RPN)生成指示灯的候选框ROI(Region of interest)，其中包括底座的候选框以及点亮状态的指示灯的候选框。而后还可以利用池化层可以得到固定尺寸的候选框的特征图。例如对于每一个ROI，将特征图的尺寸缩放到7*7，而后再通过分类模块进行N+2类的类别的判定(增加了background背景类别)，得到输入图像中每个目标对象的候选框的预测类型和位置。而后进行NMS和阈值等后处理，得到目标对象最终的检测框(目标区域对应的候选框)。
其中，本公开实施例将检测的目标对象中将点亮状态的指示灯分成N类的合理性说明：
1、不同种类的点亮状态的指示灯的意义并不相同，往往需要分别研究各个类型的检测结果，比 方说行人灯和车辆圆斑灯不能混为一谈。
2、不同种类的点亮状态的指示灯之间存在严重的样本数量不均衡的问题，将点亮状态的指示灯细分成N个不同的类别，方便调节模型的参数，分别调节优化。
在得到每个目标对象的检测结果的情况下，可以进一步识别目标对象的指示信息。其中可以通过匹配的分类器对指示信息进行分类识别。其中可以利用包括多个分类器的分类模块执行对目标对象的指示信息的识别。其中，分类模块可以包括多种类型的分类器，用于执行不同指示信息的分类识别，或者还可以包括用于提取特征的卷积层，本公开对此不作具体限定。
分类模块的输入可以为检测得到的目标对象的目标区域对应的图像特征，输出为目标区域的目标对象各自对应的指示信息。
具体过程可以包括：输入目标对象的目标区域的检测框，选择检测框内目标对象的类型(1到N+1)匹配的分类器，得到相应的分类结果。如果是指示灯底座的检测框，由于指示灯底座可以视作一个简单的整体，那么指示灯底座的分类器全部激活，例如用于识别场景和排列方式的分类器全部被激活，用以识别场景属性和排列方式属性；如果是点亮状态的指示灯的检测框，则不同类型的点亮状态的指示灯需选择不同的分类器，例如，箭头灯对应着“颜色”和“箭头方向”两个分类器，圆斑灯对应着“颜色”分类器，等等。另外，如果增加了其他属性判定的需求，也可以再增加其他的分类器，本公开对此不作具体限定。
综上所述，本公开实施例可以首先对输入图像进行目标检测处理，得到目标对象检测结果，其中目标对象的检测结果可以包括目标对象的位置以及类型等信息，再进一步根据目标对象的检测结果执行目标对象的指示信息的识别。
本公开通过将目标对象的检测过程，划分为对底座和点亮状态的指示灯这两个检测过程，在检测过程中实现了对目标对象的首次区分，后续基于目标对象的检测结果进行进一步识别时，有利于降低在识别目标对象的指示信息的过程中的识别复杂度，降低识别难度，可以简单方便的实现不同情况下的对各类型的指示灯的检测识别。
另外，本公开实施例仅仅单纯使用图片信息而不使用其他传感器，实现指示灯的检测和指示信息判定，同时本公开实施例可以针对不同类型的指示灯进行检测，具有更好的适用性。
图9示出根据本公开实施例的一种驾驶控制方法的流程图，该驾驶控制方法可以应用在智能车辆、智能飞行器、玩具等可以根据控制指令调节驾驶参数的设备中。所述驾驶控制方法可以包括：
S100：利用智能驾驶设备中的图像采集设备采集行驶图像；
在智能驾驶设备的行驶过程中，可以设置在智能驾驶设备中的图像采集设备采集行驶图像，或者也可以接收来自其他设备采集的行驶位置处的形式图像。
S200：对所述行驶图像执行所述的指示灯指示信息识别方法，得到针对所述行驶图像的指示信息；
对该行驶图像执行指示信息的检测处理，即执行上述实施例所述的指示灯指示信息的识别方法，得到行驶图像中指示灯的指示信息。
S300：利用所述指示信息生成智能驾驶设备的控制指令。
基于得到的指示信息可以实时的控制驾驶设备的驾驶参数，即可以根据得到的指示信息生成用于控制职能驾驶设备的控制指令，该控制指令可以用于控制智能驾驶设备的驾驶参数，驾驶参数可以包括驾驶速度、驾驶方向、驾驶模式或者驾驶状态中的至少一种。对于驾驶设备的参数控制或者控制指令的类型，本领域技术人员可以根据现有技术手段以及需求进行设定，本公开对此不作具体限定。
基于本公开实施例，可以实现智能驾驶设备的智能控制，由于指示信息的获取过程具有简单快速，且精度高的特点，可以提高只能驾驶设备的控制效率和精度。
本领域技术人员可以理解，在具体实施方式的上述方法中，各步骤的撰写顺序并不意味着严格的执行顺序而对实施过程构成任何限定，各步骤的具体执行顺序应当以其功能和可能的内在逻辑确定。在不违背逻辑的情况下，本公开提供的不同实现方式之间可以相互结合。
可以理解，本公开提及的上述各个方法实施例，在不违背原理逻辑的情况下，均可以彼此相互结合形成结合后的实施例，限于篇幅，本公开不再赘述。
此外，本公开还提供了指示灯的指示信息识别装置、驾驶控制装置、电子设备、计算机可读存储介质、程序，上述均可用来实现本公开提供的任一种指示灯的指示信息识别方法和/或驾驶控制方法，相应技术方案和描述和参见方法部分的相应记载，不再赘述。
图10示出根据本公开实施例的一种指示灯的指示信息识别装置的框图，如图10所示，所述指示灯的指示信息识别装置包括：
获取模块10，其用于获取输入图像；
检测模块20，其用于基于所述输入图像，确定目标对象的检测结果，所述目标对象包括指示灯底座、点亮状态的指示灯中的至少一种，所述检测结果包括所述目标对象的类型、所述输入图像中所述目标对象所在的目标区域的位置；
识别模块30，其用于基于所述目标对象的检测结果，对所述输入图像中所述目标对象所在的目标区域进行识别，得到所述目标对象的指示信息。
在一些可能的实施方式中，所述确定模块还用于：
提取所述输入图像的图像特征；
基于所述输入图像的图像特征，确定所述目标对象的至少一个候选区域中每个候选区域的第一位置；
基于所述输入图像中每个候选区域对应的第一位置处的图像特征，确定每个候选区域的中间检测结果，所述中间检测结果包括所述目标对象的预测类型和所述目标对象为所述预测类型的预测概率；所述预测类型为指示灯底座、N种点亮状态的指示灯中的任一种，N为正整数；
基于至少一个候选区域中每个候选区域的中间检测结果以及每个候选区域的第一位置，确定所述目标对象的检测结果。
在一些可能的实施方式中，所述确定模块还用于：针对每个候选区域，基于该候选区域对应的第一位置处的图像特征，对该候选区域内的所述目标对象进行分类，得到所述目标对象为所述至少一种预设类型中每种预设类型的预测概率；其中，所述预设类型包括指示灯底座、N种点亮状态的指示灯中的至少一种，N为正整数；
将所述至少一种预设类型中预测概率最高的预设类型，作为该候选区域内所述目标对象的预测类型，并得到所述预测类型的预测概率。
在一些可能的实施方式中，所述确定模块还用于：在基于至少一个候选区域中每个候选区域的中间检测结果以及每个候选区域的第一位置，确定所述目标对象的检测结果之前，基于所述输入图像的图像特征，确定每个候选区域的第一位置的位置偏差；
利用每个候选区域对应的位置偏差，调整每个候选区域的第一位置。
在一些可能的实施方式中，所述确定模块还用于在所述目标对象的候选区域为至少两个的情况下，基于至少两个候选区域中每个候选区域的中间检测结果，或者，基于每个候选区域的中间检测结果以及每个候选区域的第一位置，从所述至少两个候选区域中筛选出目标区域；
将所述目标区域内所述目标对象的预测类型作为所述目标对象的类型，将所述目标区域的第一位置作为所述目标对象所在的目标区域的位置，得到所述目标对象的检测结果。
在一些可能的实施方式中，所述确定模块还用于在所述目标对象的检测结果中仅包括指示灯底座对应的检测结果的情况下，确定所述指示灯为故障状态；
在所述目标对象的检测结果中仅包括点亮状态的指示灯对应的检测结果的情况下，确定采集所述输入图像的场景状态为黑暗状态。
在一些可能的实施方式中，所述识别模块还用于基于所述目标对象的检测结果中所述目标对象的类型，确定与所述目标对象匹配的分类器；
利用匹配的分类器，对所述输入图像中所述目标区域的图像特征进行识别，得到所述目标对象的指示信息。
在一些可能的实施方式中，所述识别模块还用于在所述目标对象的类型为指示灯底座的情况下，确定匹配的分类器包括用于对所述指示灯底座中的指示灯的排列方式进行识别的第一分类器，利用所 述第一分类器，对所述目标对象所在的目标区域的图像特征进行识别，确定所述指示灯底座中的指示灯的排列方式；和/或，
确定匹配的分类器包括用于对所述指示灯所在场景进行识别的第二分类器，利用所述第二分类器，对所述目标对象所在的目标区域的图像特征进行识别，确定所述指示灯所在的场景信息。
在一些可能的实施方式中，所述识别模块还用于在所述目标对象的类型为圆斑灯或者行人灯的情况下，确定匹配的分类器包括用于对圆斑灯的颜色属性进行识别的第三分类器；
利用所述第三分类器，对所述目标对象所在的目标区域的图像特征进行识别，确定所述圆斑灯或者行人灯的颜色属性。
在一些可能的实施方式中，所述识别模块还用于在所述目标对象的类型为箭头灯的情况下，确定匹配的分类器包括用于对箭头灯的颜色属性的第四分类器以及方向属性进行识别的第五分类器；
利用所述第四分类器和所述第五分类器，对所述目标对象所在的目标区域的图像特征进行识别，分别确定所述箭头灯的颜色属性和方向属性。
在一些可能的实施方式中，所述识别模块还用于在所述目标对象的类型为数字灯的情况下，确定匹配的分类器包括用于对数字灯的颜色属性的第六分类器以及数值属性进行识别的第七分类器；
基于所述第六分类器和所述第七分类器，对所述目标对象所在的目标区域的图像特征进行识别，分别确定所述数字灯的颜色属性和数值属性。
在一些可能的实施方式中，所述装置还包括匹配模块，其用于在所述输入图像中包括至少两个指示灯底座的情况下，针对第一指示灯底座，确定与所述第一指示灯底座匹配的点亮状态的指示灯；第一指示灯底座为所述至少两个指示灯底座中之一；
将所述第一指示灯底座的指示信息、与所述第一指示灯底座匹配的点亮状态的指示灯的指示信息进行组合，得到组合后的指示信息。
在一些可能的实施方式中，所述匹配模块还用于：
基于所述目标对象的检测结果中目标对象所在的目标区域的位置，确定所述至少一个点亮状态的指示灯所在的目标区域与所述第一指示灯底座所在的目标区域之间相交的第一面积，以及，确定所述至少一个点亮状态的指示灯所在的目标区域的第二面积；
在存在点亮状态的第一指示灯和所述第一指示灯底座之间的所述第一面积，与所述点亮状态的第一指示灯的第二面积之间的比值大于设定面积阈值的情况下，确定所述点亮状态的第一指示灯与所述第一指示灯底座相匹配；
其中，所述点亮状态的第一指示灯为所述至少一个点亮状态的指示灯中之一。
另外，图11示出根据本公开实施例的一种驾驶控制装置的框图；所述驾驶控制装置包括：
图像采集模块100，其设置在智能驾驶设备中，并用于采集所述智能驾驶设备的行驶图像；
图像处理模块200，其用于对所述行驶图像执行如第一方面中任意一项所述的指示灯指示信息识别方法，得到针对所述行驶图像的指示信息；
控制模块300，其用于利用所述指示信息，生成控制智能驾驶设备的控制指令。
在一些实施例中，本公开实施例提供的装置具有的功能或包含的模块可以用于执行上文方法实施例描述的方法，其具体实现可以参照上文方法实施例的描述，为了简洁，这里不再赘述。
本公开实施例还提出一种计算机可读存储介质，其上存储有计算机程序指令，所述计算机程序指令被处理器执行时实现上述方法。计算机可读存储介质可以是非易失性计算机可读存储介质或易失性计算机可读存储介质。
本公开实施例还提出一种电子设备，包括：处理器；用于存储处理器可执行指令的存储器；其中，所述处理器被配置为上述方法。
本公开实施例还提出一种计算机程序，包括计算机可读代码，当所述计算机可读代码在电子设备中运行时，所述电子设备中的处理器执行用于实现上述方法。
电子设备可以被提供为终端、服务器或其它形态的设备。
图12示出根据本公开实施例的一种电子设备的框图。例如，电子设备800可以是移动电话，计算 机，数字广播终端，消息收发设备，游戏控制台，平板设备，医疗设备，健身设备，个人数字助理等终端。
参照图12，电子设备800可以包括以下一个或多个组件：处理组件802，存储器804，电源组件806，多媒体组件808，音频组件810，输入/输出(I/O)的接口812，传感器组件814，以及通信组件816。
处理组件802通常控制电子设备800的整体操作，诸如与显示，电话呼叫，数据通信，相机操作和记录操作相关联的操作。处理组件802可以包括一个或多个处理器820来执行指令，以完成上述的方法的全部或部分步骤。此外，处理组件802可以包括一个或多个模块，便于处理组件802和其他组件之间的交互。例如，处理组件802可以包括多媒体模块，以方便多媒体组件808和处理组件802之间的交互。
存储器804被配置为存储各种类型的数据以支持在电子设备800的操作。这些数据的示例包括用于在电子设备800上操作的任何应用程序或方法的指令，联系人数据，电话簿数据，消息，图片，视频等。存储器804可以由任何类型的易失性或非易失性存储设备或者它们的组合实现，如静态随机存取存储器(SRAM)，电可擦除可编程只读存储器(EEPROM)，可擦除可编程只读存储器(EPROM)，可编程只读存储器(PROM)，只读存储器(ROM)，磁存储器，快闪存储器，磁盘或光盘。
电源组件806为电子设备800的各种组件提供电力。电源组件806可以包括电源管理系统，一个或多个电源，及其他与为电子设备800生成、管理和分配电力相关联的组件。
多媒体组件808包括在所述电子设备800和用户之间的提供一个输出接口的屏幕。在一些实施例中，屏幕可以包括液晶显示器(LCD)和触摸面板(TP)。如果屏幕包括触摸面板，屏幕可以被实现为触摸屏，以接收来自用户的输入信号。触摸面板包括一个或多个触摸传感器以感测触摸、滑动和触摸面板上的手势。所述触摸传感器可以不仅感测触摸或滑动动作的边界，而且还检测与所述触摸或滑动操作相关的持续时间和压力。在一些实施例中，多媒体组件808包括一个前置摄像头和/或后置摄像头。当电子设备800处于操作模式，如拍摄模式或视频模式时，前置摄像头和/或后置摄像头可以接收外部的多媒体数据。每个前置摄像头和后置摄像头可以是一个固定的光学透镜系统或具有焦距和光学变焦能力。
音频组件810被配置为输出和/或输入音频信号。例如，音频组件810包括一个麦克风(MIC)，当电子设备800处于操作模式，如呼叫模式、记录模式和语音识别模式时，麦克风被配置为接收外部音频信号。所接收的音频信号可以被进一步存储在存储器804或经由通信组件816发送。在一些实施例中，音频组件810还包括一个扬声器，用于输出音频信号。
I/O接口812为处理组件802和外围接口模块之间提供接口，上述外围接口模块可以是键盘，点击轮，按钮等。这些按钮可包括但不限于：主页按钮、音量按钮、启动按钮和锁定按钮。
传感器组件814包括一个或多个传感器，用于为电子设备800提供各个方面的状态评估。例如，传感器组件814可以检测到电子设备800的打开/关闭状态，组件的相对定位，例如所述组件为电子设备800的显示器和小键盘，传感器组件814还可以检测电子设备800或电子设备800一个组件的位置改变，用户与电子设备800接触的存在或不存在，电子设备800方位或加速/减速和电子设备800的温度变化。传感器组件814可以包括接近传感器，被配置用来在没有任何的物理接触时检测附近物体的存在。传感器组件814还可以包括光传感器，如CMOS或CCD图像传感器，用于在成像应用中使用。在一些实施例中，该传感器组件814还可以包括加速度传感器，陀螺仪传感器，磁传感器，压力传感器或温度传感器。
通信组件816被配置为便于电子设备800和其他设备之间有线或无线方式的通信。电子设备800可以接入基于通信标准的无线网络，如WiFi，2G或3G，或它们的组合。在一个示例性实施例中，通信组件816经由广播信道接收来自外部广播管理系统的广播信号或广播相关信息。在一个示例性实施例中，所述通信组件816还包括近场通信(NFC)模块，以促进短程通信。例如，在NFC模块可基于射频识别(RFID)技术，红外数据协会(IrDA)技术，超宽带(UWB)技术，蓝牙(BT)技术和其他技术来实现。
在示例性实施例中，电子设备800可以被一个或多个应用专用集成电路(ASIC)、数字信号处理器(DSP)、数字信号处理设备(DSPD)、可编程逻辑器件(PLD)、现场可编程门阵列(FPGA)、控 制器、微控制器、微处理器或其他电子元件实现，用于执行上述方法。
在示例性实施例中，还提供了一种非易失性计算机可读存储介质或易失性计算机可读存储介质，例如包括计算机程序指令的存储器804，上述计算机程序指令可由电子设备800的处理器820执行以完成上述方法。
图13示出根据本公开实施例的一种电子设备的另一框图。例如，电子设备1900可以被提供为一服务器。参照图13，电子设备1900包括处理组件1922，其进一步包括一个或多个处理器，以及由存储器1932所代表的存储器资源，用于存储可由处理组件1922的执行的指令，例如应用程序。存储器1932中存储的应用程序可以包括一个或一个以上的每一个对应于一组指令的模块。此外，处理组件1922被配置为执行指令，以执行上述方法。
电子设备1900还可以包括一个电源组件1926被配置为执行电子设备1900的电源管理，一个有线或无线网络接口1950被配置为将电子设备1900连接到网络，和一个输入输出(I/O)接口1958。电子设备1900可以操作基于存储在存储器1932的操作系统，例如Windows ServerTM，Mac OS XTM，UnixTM,LinuxTM，FreeBSDTM或类似。
在示例性实施例中，还提供了一种非易失性计算机可读存储介质或易失性计算机可读存储介质，例如包括计算机程序指令的存储器1932，上述计算机程序指令可由电子设备1900的处理组件1922执行以完成上述方法。
本公开可以是系统、方法和/或计算机程序产品。计算机程序产品可以包括计算机可读存储介质，其上载有用于使处理器实现本公开的各个方面的计算机可读程序指令。
计算机可读存储介质可以是可以保持和存储由指令执行设备使用的指令的有形设备。计算机可读存储介质例如可以是――但不限于――电存储设备、磁存储设备、光存储设备、电磁存储设备、半导体存储设备或者上述的任意合适的组合。计算机可读存储介质的更具体的例子(非穷举的列表)包括：便携式计算机盘、硬盘、随机存取存储器(RAM)、只读存储器(ROM)、可擦式可编程只读存储器(EPROM或闪存)、静态随机存取存储器(SRAM)、便携式压缩盘只读存储器(CD-ROM)、数字多功能盘(DVD)、记忆棒、软盘、机械编码设备、例如其上存储有指令的打孔卡或凹槽内凸起结构、以及上述的任意合适的组合。这里所使用的计算机可读存储介质不被解释为瞬时信号本身，诸如无线电波或者其他自由传播的电磁波、通过波导或其他传输媒介传播的电磁波(例如，通过光纤电缆的光脉冲)、或者通过电线传输的电信号。
这里所描述的计算机可读程序指令可以从计算机可读存储介质下载到各个计算/处理设备，或者通过网络、例如因特网、局域网、广域网和/或无线网下载到外部计算机或外部存储设备。网络可以包括铜传输电缆、光纤传输、无线传输、路由器、防火墙、交换机、网关计算机和/或边缘服务器。每个计算/处理设备中的网络适配卡或者网络接口从网络接收计算机可读程序指令，并转发该计算机可读程序指令，以供存储在各个计算/处理设备中的计算机可读存储介质中。
用于执行本公开操作的计算机程序指令可以是汇编指令、指令集架构(ISA)指令、机器指令、机器相关指令、微代码、固件指令、状态设置数据、或者以一种或多种编程语言的任意组合编写的源代码或目标代码，所述编程语言包括面向对象的编程语言—诸如Smalltalk、C++等，以及常规的过程式编程语言—诸如“C”语言或类似的编程语言。计算机可读程序指令可以完全地在用户计算机上执行、部分地在用户计算机上执行、作为一个独立的软件包执行、部分在用户计算机上部分在远程计算机上执行、或者完全在远程计算机或服务器上执行。在涉及远程计算机的情形中，远程计算机可以通过任意种类的网络—包括局域网(LAN)或广域网(WAN)—连接到用户计算机，或者，可以连接到外部计算机(例如利用因特网服务提供商来通过因特网连接)。在一些实施例中，通过利用计算机可读程序指令的状态信息来个性化定制电子电路，例如可编程逻辑电路、现场可编程门阵列(FPGA)或可编程逻辑阵列(PLA)，该电子电路可以执行计算机可读程序指令，从而实现本公开的各个方面。
这里参照根据本公开实施例的方法、装置(系统)和计算机程序产品的流程图和/或框图描述了本公开的各个方面。应当理解，流程图和/或框图的每个方框以及流程图和/或框图中各方框的组合，都可以由计算机可读程序指令实现。
这些计算机可读程序指令可以提供给通用计算机、专用计算机或其它可编程数据处理装置的处理器，从而生产出一种机器，使得这些指令在通过计算机或其它可编程数据处理装置的处理器执行时，产生了实现流程图和/或框图中的一个或多个方框中规定的功能/动作的装置。也可以把这些计算机可读程序指令存储在计算机可读存储介质中，这些指令使得计算机、可编程数据处理装置和/或其他设备以特定方式工作，从而，存储有指令的计算机可读介质则包括一个制造品，其包括实现流程图和/或框图中的一个或多个方框中规定的功能/动作的各个方面的指令。
也可以把计算机可读程序指令加载到计算机、其它可编程数据处理装置、或其它设备上，使得在计算机、其它可编程数据处理装置或其它设备上执行一系列操作步骤，以产生计算机实现的过程，从而使得在计算机、其它可编程数据处理装置、或其它设备上执行的指令实现流程图和/或框图中的一个或多个方框中规定的功能/动作。
附图中的流程图和框图显示了根据本公开的多个实施例的系统、方法和计算机程序产品的可能实现的体系架构、功能和操作。在这点上，流程图或框图中的每个方框可以代表一个模块、程序段或指令的一部分，所述模块、程序段或指令的一部分包含一个或多个用于实现规定的逻辑功能的可执行指令。在有些作为替换的实现中，方框中所标注的功能也可以以不同于附图中所标注的顺序发生。例如，两个连续的方框实际上可以基本并行地执行，它们有时也可以按相反的顺序执行，这依所涉及的功能而定。也要注意的是，框图和/或流程图中的每个方框、以及框图和/或流程图中的方框的组合，可以用执行规定的功能或动作的专用的基于硬件的系统来实现，或者可以用专用硬件与计算机指令的组合来实现。
以上已经描述了本公开的各实施例，上述说明是示例性的，并非穷尽性的，并且也不限于所披露的各实施例。在不偏离所说明的各实施例的范围和精神的情况下，对于本技术领域的普通技术人员来说许多修改和变更都是显而易见的。本文中所用术语的选择，旨在最好地解释各实施例的原理、实际应用或对市场中技术的技术改进，或者使本技术领域的其它普通技术人员能理解本文披露的各实施例。

Claims (31)

  1. 一种指示灯的指示信息识别方法，其特征在于，包括：
    获取输入图像；
    基于所述输入图像，确定目标对象的检测结果，所述目标对象包括指示灯底座、点亮状态的指示灯中的至少一种，所述检测结果包括所述目标对象的类型、所述输入图像中所述目标对象所在的目标区域的位置；
    基于所述目标对象的检测结果，对所述输入图像中所述目标对象所在的目标区域进行识别，得到所述目标对象的指示信息。
  2. 根据权利要求1所述的方法，其特征在于，所述基于所述输入图像，确定目标对象的检测结果，包括：
    提取所述输入图像的图像特征；
    基于所述输入图像的图像特征，确定所述目标对象的至少一个候选区域中每个候选区域的第一位置；
    基于所述输入图像中每个候选区域对应的第一位置处的图像特征，确定每个候选区域的中间检测结果，所述中间检测结果包括所述目标对象的预测类型和所述目标对象为所述预测类型的预测概率；所述预测类型为指示灯底座、N种点亮状态的指示灯中的任一种，N为正整数；
    基于至少一个候选区域中每个候选区域的中间检测结果以及每个候选区域的第一位置，确定所述目标对象的检测结果。
  3. 根据权利要求2所述的方法，其特征在于，基于所述输入图像中每个候选区域对应的第一位置处的图像特征，确定每个候选区域的中间检测结果，包括：
    针对每个候选区域，基于该候选区域对应的第一位置处的图像特征，对该候选区域内的所述目标对象进行分类，得到所述目标对象为所述至少一种预设类型中每种预设类型的预测概率；其中，所述预设类型包括指示灯底座、N种点亮状态的指示灯中的至少一种，N为正整数；
    将所述至少一种预设类型中预测概率最高的预设类型，作为该候选区域内所述目标对象的预测类型，并得到所述预测类型的预测概率。
  4. 根据权利要求2或3所述的方法，其特征在于，在基于至少一个候选区域中每个候选区域的中间检测结果以及每个候选区域的第一位置，确定所述目标对象的检测结果之前，还包括：
    基于所述输入图像的图像特征，确定每个候选区域的第一位置的位置偏差；
    利用每个候选区域对应的位置偏差，调整每个候选区域的第一位置。
  5. 根据权利要求2至4任意一项所述的方法，其特征在于，所述基于至少一个候选区域中每个候选区域的中间检测结果以及每个候选区域的第一位置，确定所述目标对象的检测结果，包括：
    响应于所述目标对象的候选区域为至少两个，基于至少两个候选区域中每个候选区域的中间检测结果，或者，基于每个候选区域的中间检测结果以及每个候选区域的第一位置，从所述至少两个候选区域中筛选出目标区域；
    将所述目标区域内所述目标对象的预测类型作为所述目标对象的类型，将所述目标区域的第一位置作为所述目标对象所在的目标区域的位置，得到所述目标对象的检测结果。
  6. 根据权利要求1至5中任意一项所述的方法，其特征在于，所述基于所述输入图像，确定目标对象的检测结果之后，所述方法还包括以下至少一种：
    响应于所述目标对象的检测结果中仅包括指示灯底座对应的检测结果，确定所述指示灯为故障状态；
    响应于所述目标对象的检测结果中仅包括点亮状态的指示灯对应的检测结果，确定采集所述输入图像的场景状态为黑暗状态。
  7. 根据权利要求1至6中任意一项所述的方法，其特征在于，所述基于所述目标对象的检测结果，对所述输入图像中所述目标对象所在的目标区域进行识别，得到所述目标对象的指示信息，包括：
    基于所述目标对象的检测结果中所述目标对象的类型，确定与所述目标对象匹配的分类器；
    利用匹配的分类器，对所述输入图像中所述目标区域的图像特征进行识别，得到所述目标对象的 指示信息。
  8. 根据权利要求7所述的方法，其特征在于，所述基于所述目标对象的检测结果，对所述输入图像中所述目标对象所在的目标区域进行识别，得到所述目标对象的指示信息，包括：
    响应于所述目标对象的类型为指示灯底座，确定匹配的分类器包括用于对所述指示灯底座中的指示灯的排列方式进行识别的第一分类器，利用所述第一分类器，对所述目标对象所在的目标区域的图像特征进行识别，确定所述指示灯底座中的指示灯的排列方式；和/或，
    确定匹配的分类器包括用于对所述指示灯所在场景进行识别的第二分类器，利用所述第二分类器，对所述目标对象所在的目标区域的图像特征进行识别，确定所述指示灯所在的场景信息。
  9. 根据权利要求7或8所述的方法，其特征在于，所述基于所述目标对象的检测结果，对所述输入图像中所述目标对象所在的目标区域进行识别，得到所述目标对象的指示信息，包括：
    响应于所述目标对象的类型为圆斑灯或者行人灯，确定匹配的分类器包括用于对圆斑灯的颜色属性进行识别的第三分类器；
    利用所述第三分类器，对所述目标对象所在的目标区域的图像特征进行识别，确定所述圆斑灯或者行人灯的颜色属性。
  10. 根据权利要求7至9中任意一项所述的方法，其特征在于，所述基于所述目标对象的检测结果，对所述输入图像中所述目标对象所在的目标区域进行识别，得到所述目标对象的指示信息，包括：
    响应于所述目标对象的类型为箭头灯，确定匹配的分类器包括用于对箭头灯的颜色属性的第四分类器以及方向属性进行识别的第五分类器；
    利用所述第四分类器和所述第五分类器，对所述目标对象所在的目标区域的图像特征进行识别，分别确定所述箭头灯的颜色属性和方向属性。
  11. 根据权利要求7至10中任意一项所述的方法，其特征在于，所述基于所述目标对象的检测结果，对所述输入图像中所述目标对象所在的目标区域进行识别，得到所述目标对象的指示信息，包括：
    响应于所述目标对象的类型为数字灯，确定匹配的分类器包括用于对数字灯的颜色属性的第六分类器以及数值属性进行识别的第七分类器；
    基于所述第六分类器和所述第七分类器，对所述目标对象所在的目标区域的图像特征进行识别，分别确定所述数字灯的颜色属性和数值属性。
  12. 根据权利要求1至11中任意一项所述的方法，其特征在于，响应于所述输入图像中包括至少两个指示灯底座，所述方法还包括：
    针对第一指示灯底座，确定与所述第一指示灯底座匹配的点亮状态的指示灯；第一指示灯底座为所述至少两个指示灯底座中之一；
    将所述第一指示灯底座的指示信息、与所述第一指示灯底座匹配的点亮状态的指示灯的指示信息进行组合，得到组合后的指示信息。
  13. 根据权利要求12所述的方法，其特征在于，所述确定与所述第一指示灯底座匹配的点亮状态的指示灯，包括：
    基于所述目标对象的检测结果中目标对象所在的目标区域的位置，确定所述至少一个点亮状态的指示灯所在的目标区域与所述第一指示灯底座所在的目标区域之间相交的第一面积，以及，确定所述至少一个点亮状态的指示灯所在的目标区域的第二面积；
    响应于存在点亮状态的第一指示灯和所述第一指示灯底座之间的所述第一面积，与所述点亮状态的第一指示灯的第二面积之间的比值大于设定面积阈值，确定所述点亮状态的第一指示灯与所述第一指示灯底座相匹配；
    其中，所述点亮状态的第一指示灯为所述至少一个点亮状态的指示灯中之一。
  14. 一种驾驶控制方法，其特征在于，包括：
    利用智能驾驶设备中的图像采集设备采集行驶图像；
    对所述行驶图像执行如权利要求1至13中任意一项所述的指示灯指示信息识别方法，得到针对所述行驶图像的指示信息；
    利用所述指示信息，生成所述智能驾驶设备的控制指令。
  15. 一种指示灯的指示信息识别装置，其特征在于，包括：
    获取模块，其用于获取输入图像；
    检测模块，其用于基于所述输入图像，确定目标对象的检测结果，所述目标对象包括指示灯底座、点亮状态的指示灯中的至少一种，所述检测结果包括所述目标对象的类型、所述输入图像中所述目标对象所在的目标区域的位置；
    识别模块，其用于基于所述目标对象的检测结果，对所述输入图像中所述目标对象所在的目标区域进行识别，得到所述目标对象的指示信息。
  16. 根据权利要求15所述的装置，其特征在于，所述确定模块还用于：
    提取所述输入图像的图像特征；
    基于所述输入图像的图像特征，确定所述目标对象的至少一个候选区域中每个候选区域的第一位置；
    基于所述输入图像中每个候选区域对应的第一位置处的图像特征，确定每个候选区域的中间检测结果，所述中间检测结果包括所述目标对象的预测类型和所述目标对象为所述预测类型的预测概率；所述预测类型为指示灯底座、N种点亮状态的指示灯中的任一种，N为正整数；
    基于至少一个候选区域中每个候选区域的中间检测结果以及每个候选区域的第一位置，确定所述目标对象的检测结果。
  17. 根据权利要求16所述的装置，其特征在于，所述确定模块还用于：针对每个候选区域，基于该候选区域对应的第一位置处的图像特征，对该候选区域内的所述目标对象进行分类，得到所述目标对象为所述至少一种预设类型中每种预设类型的预测概率；其中，所述预设类型包括指示灯底座、N种点亮状态的指示灯中的至少一种，N为正整数；
    将所述至少一种预设类型中预测概率最高的预设类型，作为该候选区域内所述目标对象的预测类型，并得到所述预测类型的预测概率。
  18. 根据权利要求16或17所述的装置，其特征在于，所述确定模块还用于：在基于至少一个候选区域中每个候选区域的中间检测结果以及每个候选区域的第一位置，确定所述目标对象的检测结果之前，基于所述输入图像的图像特征，确定每个候选区域的第一位置的位置偏差；
    利用每个候选区域对应的位置偏差，调整每个候选区域的第一位置。
  19. 根据权利要求16-18中任意一项所述的装置，其特征在于，所述确定模块还用于在所述目标对象的候选区域为至少两个的情况下，基于至少两个候选区域中每个候选区域的中间检测结果，或者，基于每个候选区域的中间检测结果以及每个候选区域的第一位置，从所述至少两个候选区域中筛选出目标区域；
    将所述目标区域内所述目标对象的预测类型作为所述目标对象的类型，将所述目标区域的第一位置作为所述目标对象所在的目标区域的位置，得到所述目标对象的检测结果。
  20. 根据权利要求15至19中任意一项所述的装置，其特征在于，所述确定模块还用于在所述目标对象的检测结果中仅包括指示灯底座对应的检测结果的情况下，确定所述指示灯为故障状态；
    在所述目标对象的检测结果中仅包括点亮状态的指示灯对应的检测结果的情况下，确定采集所述输入图像的场景状态为黑暗状态。
  21. 根据权利要求15至20中任意一项所述的装置，其特征在于，所述识别模块还用于基于所述目标对象的检测结果中所述目标对象的类型，确定与所述目标对象匹配的分类器；
    利用匹配的分类器，对所述输入图像中所述目标区域的图像特征进行识别，得到所述目标对象的指示信息。
  22. 根据权利要求21所述的装置，其特征在于，所述识别模块还用于在所述目标对象的类型为指示灯底座的情况下，确定匹配的分类器包括用于对所述指示灯底座中的指示灯的排列方式进行识别的第一分类器，利用所述第一分类器，对所述目标对象所在的目标区域的图像特征进行识别，确定所述指示灯底座中的指示灯的排列方式；和/或，
    确定匹配的分类器包括用于对所述指示灯所在场景进行识别的第二分类器，利用所述第二分类器，对所述目标对象所在的目标区域的图像特征进行识别，确定所述指示灯所在的场景信息。
  23. 根据权利要求21或22所述的装置，其特征在于，所述识别模块还用于在所述目标对象的类型为圆斑灯或者行人灯的情况下，确定匹配的分类器包括用于对圆斑灯的颜色属性进行识别的第三分类器；
    利用所述第三分类器，对所述目标对象所在的目标区域的图像特征进行识别，确定所述圆斑灯或者行人灯的颜色属性。
  24. 根据权利要求21至23中任意一项所述的装置，其特征在于，所述识别模块还用于在所述目标对象的类型为箭头灯的情况下，确定匹配的分类器包括用于对箭头灯的颜色属性的第四分类器以及方向属性进行识别的第五分类器；
    利用所述第四分类器和所述第五分类器，对所述目标对象所在的目标区域的图像特征进行识别，分别确定所述箭头灯的颜色属性和方向属性。
  25. 根据权利要求21至24中任意一项所述的装置，其特征在于，所述识别模块还用于在所述目标对象的类型为数字灯的情况下，确定匹配的分类器包括用于对数字灯的颜色属性的第六分类器以及数值属性进行识别的第七分类器；
    基于所述第六分类器和所述第七分类器，对所述目标对象所在的目标区域的图像特征进行识别，分别确定所述数字灯的颜色属性和数值属性。
  26. 根据权利要求15至25中任意一项所述的装置，其特征在于，所述装置还包括匹配模块，其用于在所述输入图像中包括至少两个指示灯底座的情况下，针对第一指示灯底座，确定与所述第一指示灯底座匹配的点亮状态的指示灯；第一指示灯底座为所述至少两个指示灯底座中之一；
    将所述第一指示灯底座的指示信息、与所述第一指示灯底座匹配的点亮状态的指示灯的指示信息进行组合，得到组合后的指示信息。
  27. 根据权利要求26所述的装置，其特征在于，所述匹配模块还用于：
    基于所述目标对象的检测结果中目标对象所在的目标区域的位置，确定所述至少一个点亮状态的指示灯所在的目标区域与所述第一指示灯底座所在的目标区域之间相交的第一面积，以及，确定所述至少一个点亮状态的指示灯所在的目标区域的第二面积；
    在存在点亮状态的第一指示灯和所述第一指示灯底座之间的所述第一面积，与所述点亮状态的第一指示灯的第二面积之间的比值大于设定面积阈值的情况下，确定所述点亮状态的第一指示灯与所述第一指示灯底座相匹配；
    其中，所述点亮状态的第一指示灯为所述至少一个点亮状态的指示灯中之一。
  28. 一种驾驶控制装置，其特征在于，包括：
    图像采集模块，其设置在智能驾驶设备中，并用于采集所述智能驾驶设备的行驶图像；
    图像处理模块，其用于对所述行驶图像执行如权利要求1至13中任意一项所述的指示灯指示信息识别方法，得到针对所述行驶图像的指示信息；
    控制模块，其用于利用所述指示信息，生成所述智能驾驶设备的控制指令。
  29. 一种电子设备，其特征在于，包括：
    处理器；
    用于存储处理器可执行指令的存储器；
    其中，所述处理器被配置为调用所述存储器存储的指令，以执行权利要求1至13中任意一项所述的方法，或者执行权利要求14所述的方法。
  30. 一种计算机可读存储介质，其上存储有计算机程序指令，其特征在于，所述计算机程序指令被处理器执行时实现权利要求1至13中任意一项所述的方法，或者实现权利要求14所述的方法。
  31. 一种计算机程序，包括计算机可读代码，其特征在于，当所述计算机可读代码在电子设备中运行时，所述电子设备中的处理器执行用于实现权利要求1至13中任意一项所述的方法，或者实现权利要求14所述的方法。
PCT/CN2020/095437 2019-06-27 2020-06-10 指示灯的指示信息识别方法及装置、电子设备和存储介质 WO2020259291A1 (zh)

Priority Applications (4)

Application Number Priority Date Filing Date Title
SG11202102205TA SG11202102205TA (en) 2019-06-27 2020-06-10 Indication information identification method and apparatus for indicator light, and electronic device and storage medium
JP2021512798A JP2022500739A (ja) 2019-06-27 2020-06-10 表示灯の指示情報認識方法及び装置、電子機器並びに記憶媒体
KR1020217009669A KR20210052525A (ko) 2019-06-27 2020-06-10 표시등의 지시 정보 인식 방법 및 장치, 전자 기기 및 기억 매체
US17/194,175 US20210192239A1 (en) 2019-06-27 2021-03-05 Method for recognizing indication information of an indicator light, electronic apparatus and storage medium

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910569896.8A CN112149697A (zh) 2019-06-27 2019-06-27 指示灯的指示信息识别方法及装置、电子设备和存储介质
CN201910569896.8 2019-06-27

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US17/194,175 Continuation US20210192239A1 (en) 2019-06-27 2021-03-05 Method for recognizing indication information of an indicator light, electronic apparatus and storage medium

Publications (1)

Publication Number Publication Date
WO2020259291A1 true WO2020259291A1 (zh) 2020-12-30

Family

ID=73868880

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/095437 WO2020259291A1 (zh) 2019-06-27 2020-06-10 指示灯的指示信息识别方法及装置、电子设备和存储介质

Country Status (6)

Country Link
US (1) US20210192239A1 (zh)
JP (1) JP2022500739A (zh)
KR (1) KR20210052525A (zh)
CN (1) CN112149697A (zh)
SG (1) SG11202102205TA (zh)
WO (1) WO2020259291A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114821194A (zh) * 2022-05-30 2022-07-29 深圳市科荣软件股份有限公司 一种设备运行状态识别方法及装置

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP7172441B2 (ja) * 2018-10-25 2022-11-16 トヨタ自動車株式会社 進行可能方向検出装置及び進行可能方向検出方法
CN112712057B (zh) * 2021-01-13 2021-12-07 腾讯科技（深圳）有限公司 交通信号识别方法、装置、电子设备及存储介质
CN113138887A (zh) * 2021-04-25 2021-07-20 浪潮商用机器有限公司 一种服务器故障灯检测方法、装置及系统
CN113408409A (zh) * 2021-06-17 2021-09-17 阿波罗智联(北京)科技有限公司 交通信号灯识别方法、设备、云控平台和车路协同系统
CN113269190B (zh) * 2021-07-21 2021-10-12 中国平安人寿保险股份有限公司 基于人工智能的数据分类方法、装置、计算机设备及介质
CN113808117B (zh) * 2021-09-24 2024-05-21 北京市商汤科技开发有限公司 灯具检测方法、装置、设备及存储介质
CN115214430B (zh) * 2022-03-23 2023-11-17 广州汽车集团股份有限公司 车辆座椅调节方法及车辆

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100271497A1 (en) * 2009-04-28 2010-10-28 Monsive Jr Michael G Portable traffic monitoring system and methods for use
CN102176287A (zh) * 2011-02-28 2011-09-07 无锡中星微电子有限公司 一种交通信号灯识别系统和方法
CN105390007A (zh) * 2015-11-17 2016-03-09 陕西科技大学 基于模式识别的交通指挥系统
CN108615383A (zh) * 2018-05-14 2018-10-02 吉林大学 一种基于车间通信的汽车交通路口辅助通行系统及其控制方法

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102117546B (zh) * 2011-03-10 2013-05-01 上海交通大学 车载交通信号灯辅助装置
DE102012108863A1 (de) * 2012-09-20 2014-05-28 Continental Teves Ag & Co. Ohg Verfahren zur Erkennung eines Ampelzustands mittels einer Kamera
US10650257B2 (en) * 2017-02-09 2020-05-12 SMR Patents S.à.r.l. Method and device for identifying the signaling state of at least one signaling device
CN107038420A (zh) * 2017-04-14 2017-08-11 北京航空航天大学 一种基于卷积网络的交通信号灯识别算法
CN108804983B (zh) * 2017-05-03 2022-03-18 腾讯科技（深圳）有限公司 交通信号灯状态识别方法、装置、车载控制终端及机动车
CN108875608B (zh) * 2018-06-05 2021-12-17 合肥湛达智能科技有限公司 一种基于深度学习的机动车交通信号识别方法
CN110660254B (zh) * 2018-06-29 2022-04-08 北京市商汤科技开发有限公司 交通信号灯检测及智能驾驶方法和装置、车辆、电子设备
CN109830114A (zh) * 2019-02-20 2019-05-31 华为技术有限公司 交通信号灯提醒方法和装置

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100271497A1 (en) * 2009-04-28 2010-10-28 Monsive Jr Michael G Portable traffic monitoring system and methods for use
CN102176287A (zh) * 2011-02-28 2011-09-07 无锡中星微电子有限公司 一种交通信号灯识别系统和方法
CN105390007A (zh) * 2015-11-17 2016-03-09 陕西科技大学 基于模式识别的交通指挥系统
CN108615383A (zh) * 2018-05-14 2018-10-02 吉林大学 一种基于车间通信的汽车交通路口辅助通行系统及其控制方法

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114821194A (zh) * 2022-05-30 2022-07-29 深圳市科荣软件股份有限公司 一种设备运行状态识别方法及装置

Also Published As

Publication number Publication date
KR20210052525A (ko) 2021-05-10
US20210192239A1 (en) 2021-06-24
SG11202102205TA (en) 2021-04-29
CN112149697A (zh) 2020-12-29
JP2022500739A (ja) 2022-01-04

Similar Documents

Publication Publication Date Title
WO2020259291A1 (zh) 指示灯的指示信息识别方法及装置、电子设备和存储介质
US11875574B2 (en) Object recognition method of autonomous driving device, and autonomous driving device
US11308809B2 (en) Collision control method and apparatus, and storage medium
CN106004883B (zh) 车辆违规提醒的方法及装置
WO2021031506A1 (zh) 图像处理方法及装置、电子设备和存储介质
US11749112B2 (en) Warning device, warning method, and warning program
WO2020238699A1 (zh) 信号指示灯的状态检测方法及装置、驾驶控制方法及装置
US9536157B2 (en) Method for identification of a projected symbol on a street in a vehicle, apparatus and vehicle
US10261170B2 (en) Image analysis and radar detectors
JP2018097738A (ja) 画像処理装置、画像処理システム、および画像処理方法
WO2021057244A1 (zh) 光强调节方法及装置、电子设备和存储介质
WO2021213241A1 (zh) 目标检测方法及装置、电子设备、存储介质及程序
EP3309711A1 (en) Vehicle alert apparatus and operating method thereof
KR20180116808A (ko) 신호등 상태 예측 장치 및 방법
KR101793156B1 (ko) 신호등을 이용한 차량 사고 방지 시스템 및 방법
CN114627443B (zh) 目标检测方法、装置、存储介质、电子设备及车辆
CN112857381A (zh) 一种路径推荐方法、装置及可读介质
CN115965935A (zh) 目标检测方法、装置、电子设备、存储介质和程序产品
CN113344900B (zh) 机场跑道侵入检测方法、装置、存储介质及电子设备
CN113129617A (zh) 行车提示方法及装置
KR20230056856A (ko) 차량에 탑재된 전자 장치 및 그 동작 방법
JP2016167174A (ja) 警告装置、警告方法及び警告プログラム
KR20210004917A (ko) 전자 장치 및 그의 제어 방법
CN114633764A (zh) 交通信号灯检测方法、装置、存储介质、电子设备及车辆
KR20200094830A (ko) 객체 인식 장치 및 방법

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20830632

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2021512798

Country of ref document: JP

Kind code of ref document: A

ENP Entry into the national phase

Ref document number: 20217009669

Country of ref document: KR

Kind code of ref document: A

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 28.03.2022)

122 Ep: pct application non-entry in european phase

Ref document number: 20830632

Country of ref document: EP

Kind code of ref document: A1