Driving violation behavior detection method with misjudgment elimination function and electronic device
By combining the joint probability density function of local and remote detection models, error values are corrected to reduce false judgments, thus solving the problems of false judgments and false alarms in traditional vehicle driving monitoring systems and improving the reliability and accuracy of detection results.
Patent Information
- Application Number
- CN202410599111.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-05-15
- Publication Date
- 2025-11-25
AI Technical Summary
Traditional vehicle driver monitoring systems, due to hardware resource limitations, often use lightweight detection models, leading to frequent misjudgments and false alarms, which affects driver trust.
The method combines a local detection model with a remote server. By using the joint probability density function of the local object detection model and the remote object detection model, the error value is calculated and compared with a preset threshold to correct the error and eliminate false judgments.
It reduces false alarms and misjudgments in the detection of driving violations, and improves the reliability and accuracy of the detection results.
Smart Images
Figure CN121010964A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a technology for detecting driving violations, and more particularly to a method and electronic device for detecting driving violations with a false positive elimination function. Background Technology
[0002] Transportation companies use Driver Monitoring Systems (DMS) to detect driver misconduct, such as drowsiness, distraction, or talking on the phone, in order to monitor and manage the driving behavior of their drivers. DMS primarily uses in-vehicle cameras to record drivers and combines this with image recognition technology to analyze their driving behavior. If a violation is detected, the incident is uploaded to the cloud.
[0003] In addition to detecting driver violations, vehicle driving monitoring systems also incorporate multi-tasking functions such as driving video recording, multi-camera image recognition, and event uploading. Due to limitations in hardware platform resources, most vehicle driving monitoring systems use lightweight detection models or algorithms for image detection and analysis.
[0004] However, traditional methods for identifying violations often rely on the threshold of object recognition in a single frame (e.g., a mobile phone) to determine whether a phone call has been made. Furthermore, using lightweight detection models or algorithms can easily fail to account for unexpected changes in the environment, leading to incorrect image recognition by the monitoring system and resulting in false alarms or missed alarms. On the other hand, it may also affect the recording of driver violations. In addition, frequent false alarms can cause drivers to lose trust in the monitoring system. Summary of the Invention
[0005] Therefore, the purpose of this invention is to provide a method for detecting driving violations with a function of eliminating false positives.
[0006] Therefore, the driving violation detection method of the present invention with false positive elimination function is applicable to detecting a local detection result detected by an object detection model indicating whether a driver has committed a driving violation while driving a vehicle. This detection method is implemented by an electronic device installed in the vehicle. The electronic device stores multiple historical images of the driver taken at past time points, a local object detection model used to determine whether the images contain multiple feature objects, and a baseline truth probability density function pre-provided by a server. The electronic device continuously captures images of the driver to continuously generate and store images of the driver. The driving violation detection method includes steps (A), (B), (C), (D), (E), and (F). In step (A), the electronic device captures images of the driver at a current time point to generate and store a current image. In step (B), the electronic device uses the local object detection model to generate multiple local historical joint probabilities corresponding to the target historical images and a local current joint probability corresponding to the current image, based on multiple target historical images taken within a predetermined time interval and the current image. In step (C), the electronic device generates a local probability density function related to the probability distribution of the driving violation occurring in the current image and the target historical images, based on the local historical joint probabilities and the local current joint probabilities. In step (D), the electronic device generates an error value based on the local probability density function and the baseline truth probability density function. In step (E), the electronic device determines whether the error value is less than a preset threshold. In step (F), when the error value is determined to be less than the preset threshold, the electronic device generates a detection result indicating that the local detection result is reliable.
[0007] Specifically, the electronic device connects to a remote server via a communication network. The remote server stores a remote object detection model used to determine whether an image contains multiple feature objects. Between steps (A) and (B), the following steps are also included: (H) The electronic device determines whether correction is needed; (I) When correction is determined to be needed, the electronic device transmits the target historical images and the current image to the remote server via the communication network. The remote server uses the remote object detection model to generate a remote detection result, and based on the target historical images and the current image, generates a remote probability density function related to the probability distribution of the driving violation occurring in the current image and the target historical images, and calculates multiple real-time parameters that conform to the remote probability density function, which are then transmitted back via the communication network; (J) The electronic device updates the baseline truth probability density function based on the real-time parameters; (K) When correction is determined to be unnecessary, step (B) is performed. 。
[0008] Specifically, in step (H), when the electronic device determines that the current time point has gone through a preset period, the electronic device determines that correction is needed.
[0009] Specifically, in step (B), the electronic device inputs the current image into the local object detection model and, based on the local detection results, multiplies the individual probability values of the feature objects in the current image when they meet the driving violation behavior to generate the local current joint probability corresponding to the current image.
[0010] Specifically, in step (B), the electronic device inputs the target historical image into the local object detection model, and outputs the individual probability values of the feature objects in each frame of the target historical image when they meet the driving violation behavior according to the image recognition results. Then, the individual probability values in each frame are multiplied together to generate multiple local historical joint probabilities corresponding to each frame of the target historical image.
[0011] Another objective of this invention is to provide an electronic device for preventing misjudgment of driving violations.
[0012] Therefore, the present invention provides an electronic device for detecting driving violations with false positive exclusion, suitable for determining whether a driver has committed a driving violation while driving a vehicle. Installed within the vehicle, the electronic device includes a storage unit, a camera unit, and a processing unit. The storage unit stores multiple historical images of the driver taken at past times, a local object detection model for determining whether an image contains multiple feature objects, and a baseline truth probability density function. The camera unit, electrically connected to the storage unit, continuously captures images of the driver to continuously generate and store images of the driver in the storage unit. The processing unit is electrically connected to the camera unit and the storage unit. The imaging unit captures a picture of the driver at a current time point to generate and store a current image in the storage unit. The processing unit uses a local object detection model to detect whether the driver has committed a violation, such as using a mobile phone to make a call. Then, based on multiple target historical images captured within a predetermined time interval in these historical images and the current image, it generates multiple local historical joint probabilities corresponding to the target historical images and a local current joint probability corresponding to the current image. Based on these local historical joint probabilities and the local current joint probabilities, the processing unit generates a local probability density function related to the probability distribution of the driving violation occurring in the current image and the target historical images. Based on the local probability density function and the baseline truth probability density function, the processing unit generates an error value. The electronic device determines whether the error value is less than a preset threshold. When the error value is determined to be less than the preset threshold, the electronic device generates a detection result indicating that the local detection result is credible.
[0013] Specifically, it also includes a communication unit electrically connected to the processing unit. The communication unit is connected to a remote server via a communication network. The remote server stores a remote object detection model for determining whether an image contains multiple feature objects. The processing unit determines whether correction is needed. When correction is needed, the electronic device transmits the target historical images and the current image to the remote server via the communication network. The remote server uses the remote object detection model to generate a remote detection result and, based on the target historical images and the current image, generates a remote probability density function related to the probability distribution of the driving violation occurring in the current image and the target historical images. It also calculates multiple real-time parameters that conform to the remote probability density function and transmits them back via the communication network. The electronic device updates the baseline truth probability density function based on these real-time parameters. When correction is not needed, the processing unit generates the local historical joint probability and the current joint probability.
[0014] Specifically, when the electronic device determines that the current time point has reached the preset period, the electronic device determines that correction is required.
[0015] Specifically, the processing unit inputs the current image into the local object detection model, which outputs multiple probability values corresponding to the individual characteristics of the objects when they meet the driving violation criteria. The processing unit then multiplies these individual probability values to generate the local current joint probability corresponding to the current image.
[0016] Specifically, the processing unit inputs the target historical image into the local object detection model, which outputs multiple probability values for each of the feature objects in each frame of the target historical image when they meet the driving violation behavior. Then, the processing unit multiplies the individual probability values in each frame to generate multiple local historical joint probabilities for each frame of the target historical image.
[0017] Compared with existing technologies, the driving violation detection method and electronic device of the present invention, which has a false judgment elimination function, examines the error value between the local probability density function and the benchmark truth probability density function, and determines whether the error value is less than a preset threshold. When the error value is determined to be less than the preset threshold, it indicates that the local detection result judged by the local object detection model is reliable. When the error value is determined to be not less than the preset threshold, it indicates that the local detection result judged by the local object detection model is unreliable. In this case, the electronic device will regenerate a new current image and re-detect and calculate. This allows for re-image recognition and calculation when there is doubt about the judgment, thereby reducing the occurrence of false judgments and false alarms in driving violation detection. Attached Figure Description
[0018] Other features and effects of the present invention will be clearly presented in the embodiments with reference to the accompanying drawings, wherein:
[0019] Figure 1 This is a block diagram illustrating an embodiment of the present invention, which has a driving violation detection electronic device for false positive exclusion.
[0020] Figure 2 This is a flowchart illustrating steps 201-206 of an embodiment of the driving violation detection method with false positive elimination function of the present invention;
[0021] Figure 3 This is a flowchart illustrating steps 207-212 of an embodiment of the driving violation detection method with false positive elimination function of the present invention; and
[0022] Figure 4It is a schematic diagram illustrating a local current joint probability value corresponding to a current image, a local joint probability distribution consisting of multiple local historical joint probability values corresponding to multiple target historical images, and a local probability density function. Detailed Implementation
[0023] Before the invention is described in detail, it should be noted that similar elements are represented by the same numbers in the following description.
[0024] See Figure 1 This invention relates to an embodiment of an electronic device 1 for implementing a driving violation detection method with a false positive elimination function. The device includes a storage unit 11, a communication unit 12, a camera unit 13, and a processing unit 14, wherein these units can transmit relevant data or data via electrical connections. The electronic device 1 is used to implement a driving violation detection method with a false positive elimination function. This function is applicable to determining whether a local detection result generated by an object detection model, indicating whether a driver has committed a violation while driving a vehicle, has resulted in a false positive.
[0025] It is particularly important to note that in this embodiment, the driving violation is, for example, driving while holding a phone and talking on it. The electronic device 1 is, for example, a vehicle driving monitoring system. In other embodiments, the driving violation may also be, for example, dozing off, and is not limited thereto. Therefore, for example, if the driver is not holding a phone while driving, but the local detection result indicates that the driver is holding a phone while driving, the aforementioned false alarm exclusion function will determine that the local detection result indicates a false alarm.
[0026] The storage unit 11 stores multiple historical images of the driver taken at past points in time, a local object detection model for determining whether an image contains multiple characteristic objects, and a baseline truth probability density function. The baseline truth probability density function can be the factory setting when the electronic device 1 leaves the factory, or it can be received from the aforementioned server when the electronic device 1 performs a calibration procedure during use and communicates with the server, and is not limited to this.
[0027] The communication unit 12 is connected to a remote server 101 via a communication network 100. The remote server 101 stores a remote object detection model for determining whether an image contains multiple feature objects.
[0028] It is worth noting that in this embodiment, both the local object detection model and the remote object detection model are built using deep learning. The local object detection model is used to generate a detection result indicating whether the driver has committed a violation while driving, based on the image of the driver and image recognition technology. The violation includes, but is not limited to, dozing off or using a mobile phone. Generally, due to considerations of vehicle interior space and driver visibility, the electronic device 1 for detecting driver violations tends to be miniaturized. Therefore, the hardware performance within the electronic device 1 is limited, and only a digital signal processor (DSP) can be used for computation. Consequently, the local object detection model can execute fewer computational layers, variables may only be in integer form, and the resolution of parameter values is poor. Conversely, compared to the electronic device 1 that can detect driving violations, the remote server 101 has fewer hardware limitations and can use a graphics processing unit (GPU) for computation. Therefore, the remote object detection model can perform more computational layers, and the variables can mostly be floating-point numbers with higher parameter resolution. Thus, the remote server 101, with its higher-level integrated computing power, uses a more complete and rigorous algorithm to perform image recognition on the remote object detection model, thereby obtaining detection results with higher reliability than the local object detection model.
[0029] It should be noted that in this embodiment, since the driving violation is committed while using a handheld phone, the features are, for example, the phone, the hand, the face, and the mouth. In other embodiments, the features may be only the phone and the hand, but are not limited thereto.
[0030] The shooting unit 13 is electrically connected to the storage unit 11 and is used to continuously shoot the driver so as to continuously generate and store the driver's images to the storage unit 11. The shooting area includes images of the driver's face and the area around the driver's seat, but is not limited thereto.
[0031] It should be noted that in this embodiment, the imaging unit 13 captures one image per second, but is not limited to this.
[0032] The processing unit 14 is electrically connected to the storage unit 11, the communication unit 12, and the imaging unit 13.
[0033] See Figure 1 , 2 Section 3 illustrates an embodiment of the driving violation detection method with false positive elimination function. The steps included in this embodiment are described in detail below.
[0034] In step 201, the camera unit 13 captures a picture of the driver at a current time point to generate and store a current image in the storage unit 11.
[0035] In step 202, the processing unit 14 determines whether a correction procedure needs to be performed. If correction is determined to be required, the process proceeds to step 203; otherwise, if correction is determined not to be required, the process proceeds to step 207.
[0036] It is particularly important to note that in this embodiment, the processing unit 14 determines whether correction needs to be performed based on the current time and a preset period. When the processing unit 14 determines that a preset period has elapsed based on the current time, it determines that a correction procedure needs to be performed, but this is not a limitation. In this embodiment, the preset period is, for example, but not limited to, 10 minutes. If calculated starting from 09:00, correction is initiated every 10 minutes at 09:10, 09:20, 09:30, etc. In other embodiments, the electronic device 1 also includes a light sensor (not shown). When the light sensor senses changes in the lighting environment, such as when a vehicle enters or leaves a tunnel, resulting in strong light and shadows covering the area around the driver's seat, the large changes in ambient light may increase the probability of errors in the image recognition results of the local object detection model. Therefore, the amount of light in the environment can be confirmed by measuring the illuminance (unit: lux) around the driver's seat. When the detected change in illuminance is greater than a set value, the processing unit 14 determines that a correction procedure needs to be performed. Furthermore, if the processing unit 14 detects that the facial features of the current image are different from those of the historical images, in order to avoid misjudgment when identifying the driving violations of different drivers due to differences in the proportions of the face shape, eyes, and mouth or differences in facial features, the processing unit 14 determines that a correction procedure needs to be performed, but this is not the only limitation.
[0037] In step 203, the processing unit 14 transmits multiple target historical images taken within a predetermined time interval in the historical images and the current image to the remote server 101 via the communication unit 12.
[0038] It is particularly important to note that in this embodiment, the predetermined time interval is from 10 seconds before the current time to 1 second before the current time. For example, if the current time is the 10th second, then the predetermined time interval is from 0 to 9 seconds, but it is not limited to this. Furthermore, if the imaging unit 13 is set to capture one image per second, then the imaging unit 13 will capture 9 images within the aforementioned predetermined time interval. The aforementioned number of images captured per second and the length of the predetermined time interval are merely illustrative examples and are not intended to be limiting.
[0039] In step 204, the remote server 101 utilizes the remote object detection model and, based on the received current image, generates a remote detection result indicating whether the driver has committed a driving violation. This remote detection result includes the occurrence of a driving violation and the probability values of individual characteristic objects when a driving violation occurs. The remote object detection model employs a complete and more rigorous algorithm to identify the current image. The remote server 101 will generate multiple remote historical joint probabilities corresponding to the received target historical images and a remote current joint probability corresponding to the current image, based on the received target historical images and the current image.
[0040] The remote historical joint probability refers to the probability of a driving violation occurring in each frame of the target historical image when the server 101 performs image recognition based on the remote object detection model to generate the remote detection result. This is achieved by multiplying the individual probability values corresponding to each feature object in each frame to generate multiple remote historical joint probabilities, representing the overall probability of a driving violation occurring in each frame of the target historical image. The remote current joint probability is generated in a similar way to the remote historical joint probabilities, except that the received image is the current image; therefore, it will not be described in detail here.
[0041] In step 205, the remote server 101 generates a remote probability density function based on the remote historical joint probability and the remote current joint probability, which relates to the probability distribution of the driving violation occurring in the target historical images and the current images within the predetermined time interval. Based on the remote probability density function, the server calculates a number of real-time parameters related to the remote probability density function and transmits these real-time parameters to the communication unit 12 via the communication network 100. The communication unit 12 then transmits these real-time parameters to the processing unit 14 via an electrical connection.
[0042] It is worth noting that, in this embodiment, the real-time parameters include at least an event-representative mean and an event-representative variance. That is, when the remote server 101 uses the remote object detection model to perform image identification on each frame of the target historical images and outputs the individual probabilities of the feature objects when they meet the driving violation behavior, the remote server 101 further generates multiple remote historical joint probabilities corresponding to each frame of the target historical images, and generates the corresponding remote current joint probabilities based on the current image. Then, it calculates the remote probability density function using the remote historical joint probabilities and the remote current joint probabilities. The remote probability density function is then used to estimate the real-time parameters that best match the result of the remote probability density function in the time interval of the complete image data for detecting the driving violation, namely the event-representative mean and the event-representative variance, but this is not a limitation.
[0043] In step 206, the processing unit 14 can update the baseline truth probability density function based on these real-time parameters.
[0044] In step 207, the processing unit 14 uses the local object detection model and the current image captured in step 201 to calculate a local detection result regarding whether the driver committed a driving violation in the current image. This local detection result also indicates whether the driver committed a driving violation, and includes the probability values of whether the driving violation occurred and the individual probability values of the characteristic objects when the driving violation is met. The processing unit 14 also generates multiple local historical joint probabilities corresponding to the target historical images and a local current joint probability corresponding to the current image, based on multiple target historical images captured within a predetermined time interval in the historical images and the current image.
[0045] It is worth noting that in steps 204 and 207 of this embodiment, both the remote server 101 and the processing unit 14 use images captured within the same predetermined time interval for identification. However, since the remote server 101 has better performance than the processing unit 14, in other embodiments, the remote server 101 can use images captured within a longer time interval than the predetermined time interval for image identification in step 204, while the processing unit 14 uses historical target images captured within the predetermined time interval for identification. That is, the remote server 101 and the processing unit 14 can also use different predetermined time intervals. The remote server 101 increases the overall accuracy of the judgment by calculating and analyzing image data over a longer period, while the processing unit 14 uses the amount of data that can meet the needs of real-time computing. For example, the remote server 101 can identify images based on 50 seconds of captured images, while the processing unit 14 can identify images based on only 10 seconds of captured images. In this embodiment, the electronic device 1 inputs the current image into the local object detection model. The processing unit 14 calculates the probability values individually corresponding to each feature object based on the local detection results of the local object detection model, multiplies them together, and generates a local current joint probability corresponding to the current image. Similarly, the processing unit 14 inputs the target historical images into the local object detection model. The processing unit 14 calculates based on the probability values individually corresponding to each feature object to generate multiple local historical joint probabilities corresponding to each frame of the target historical image.
[0046] In step 208, the processing unit 14 generates a local probability density function based on the local historical joint probabilities and the local current joint probabilities, which is related to the probability distribution of the driving violation occurring in the current image and the target historical images within the predetermined time interval.
[0047] For reference Figure 4 The example illustrates a local joint probability distribution consisting of the local current joint probability value corresponding to the current image captured at 10-11 seconds, and the local historical joint probability values corresponding to the target historical images captured at 0-10 seconds. The probability distributions of the local probability density function, the baseline truth probability density function, and the remote probability density function are, for example, the probability density function (PDF) of a normal distribution.
[0048] In step 209, the processing unit 14 generates an error value between the probability density functions based on the local probability density function and the reference truth probability density function.
[0049] It is particularly important to note that in this embodiment, the error value is the cross entropy of the baseline truth probability density function and the local probability density function. That is, the result of the cross entropy calculation is used to measure the degree of error between the electronic device 1 and the remote server 101 in detecting driving violations and their probability density distributions. If the results calculated by the two are close (high overlap), the value of the cross entropy will be lower (small difference), indicating that there is less uncertainty in the calculated data. Therefore, the reliability of the detection results of driving violations is higher, but the measurement method is not limited to this.
[0050] In step 210, the processing unit 14 determines whether the error value is less than a preset threshold. When the processing unit 14 determines that the error value is less than the preset threshold, the processing unit 14 generates a detection result indicating that the local detection result is reliable, and the process proceeds to step 211; while when the processing unit 14 determines that the error value is not less than the preset threshold, the processing unit 14 generates a detection result indicating that the local detection result is unreliable and repeats step 201.
[0051] In step 211, the processing unit 14 generates a local detection result regarding the driver's driving violation based on the local object detection model, and determines the reliability of the local detection result when the aforementioned error value is lower than the preset threshold. When the local detection result indicates that a driving violation has occurred, the process proceeds to step 212; and when the local detection result indicates that a driving violation has not occurred, the process proceeds to step 201.
[0052] In step 212, the processing unit 14 generates an alert message.
[0053] It is worth noting that the warning message can be an audio message, in which the processing unit 14 can issue the warning message through a speaker (not shown) to remind the driver to stop the distracted driving violation caused by using a mobile phone. The warning message can also be a text message, in which the processing unit 14 can transmit the warning message to the remote server 101 through the communication module, and record and store multiple driving information data such as the time, location and duration of the driver's driving violation as historical data of driver driving behavior in the fleet management function, but it is not limited to this.
[0054] In summary, the processing unit 14 of the electronic device 1 generates the error value based on the local probability density function and the baseline truth probability density function, and determines whether the error value is less than the preset threshold. When the error value is determined to be less than the preset threshold, it indicates that the local detection result of the local object detection model is reliable. When the error value is determined to be not less than the preset threshold, it indicates that the local detection result of the local object detection model is unreliable. In this case, the processing unit 14 will regenerate a new current image and recalculate the detection. This invention utilizes the computing power of the remote server 101 to execute the correction procedure, thereby obtaining real-time parameters that are more consistent with the current image data to update the baseline truth probability density function in the electronic device 1. By using the error status between the updated baseline truth probability density function and the local probability density function, some incorrect identification results of driving violations can be eliminated, thereby reducing the occurrence of misjudgments or false alarms. Therefore, the purpose of this invention can indeed be achieved.
[0055] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.
Claims
1. A method for detecting driving violations with a false positive elimination function, applicable to detecting a local detection result detected by an object detection model indicating whether a driver has committed a violation while driving a vehicle. The detection method is implemented by an electronic device installed in the vehicle. The electronic device stores multiple historical images of the driver taken at past times, a local object detection model for determining whether the images contain multiple feature objects, and a baseline truth probability density function. The electronic device continuously captures images of the driver to continuously generate and store images of the driver. The method is characterized by... This method for detecting driving violations includes the following steps: (A) The electronic device captures the driver at a current point in time to generate and store a current image; (B) The electronic device uses the local object detection model to generate multiple local historical joint probabilities corresponding to the target historical images and a local current joint probability corresponding to the current image, based on multiple target historical images taken in a predetermined time interval in the historical images and the current image. (C) The electronic device generates a local probability density function relating to the probability distribution of the driving violation occurring in the current image and the target historical images, based on the local historical joint probabilities and the local current joint probabilities. (D) The electronic device generates an error value based on the local probability density function and the reference truth probability density function; (E) The electronic device determines whether the error value is less than a preset threshold; and (F) When the error value is determined to be less than the preset threshold, the electronic device generates a detection result indicating that the local detection result is reliable.
2. The driving violation detection method with false positive elimination function according to claim 1, characterized in that, The electronic device is connected to a remote server via a communication network. The remote server stores a remote object detection model for determining whether an image contains multiple feature objects. Between steps (A) and (B), the following steps are also included: (H) The electronic device determines whether calibration is required; (I) When it is determined that correction is needed, the electronic device transmits the target historical images and the current image to the remote server via the communication network. The remote server uses the remote object detection model to generate a remote detection result, and generates a remote probability density function related to the probability distribution of the driving violation in the current image and the target historical images based on the target historical images and the current image. It also calculates multiple real-time parameters that conform to the remote probability density function and then transmits them back via the communication network. (J) The electronic device updates the baseline truth probability density function based on these real-time parameters; (K) When it is determined that no correction is needed, proceed to step (B).
3. The driving violation detection method with false positive elimination function according to claim 2, characterized in that, In step (H), when the electronic device determines that the current time point has gone through a preset period, the electronic device determines that correction is needed.
4. The driving violation detection method with false positive elimination function according to claim 1, characterized in that, In step (B), the electronic device inputs the current image into the local object detection model and, based on the local detection results, multiplies the individual probability values of the feature objects in the current image when they meet the driving violation behavior to generate the local current joint probability corresponding to the current image.
5. The driving violation detection method with false positive elimination function according to claim 1, characterized in that, In step (B), the electronic device inputs the target historical image into the local object detection model, and outputs the individual probability values of the feature objects in each frame of the target historical image when they meet the driving violation behavior according to the image recognition results. Then, the individual probability values in each frame are multiplied together to generate multiple local historical joint probabilities corresponding to each frame of the target historical image.
6. An electronic device suitable for implementing a method for detecting driving violations with a false positive elimination function, the electronic device being installed inside the vehicle, characterized in that, The electronic device includes: A storage unit stores multiple historical images of the driver taken at past points in time, a local object detection model for determining whether the images contain multiple characteristic objects, and a baseline truth probability density function. A camera unit, electrically connected to the storage unit, is used to continuously capture images of the driver in order to continuously generate and store images of the driver in the storage unit; and A processing unit is electrically connected to the imaging unit and the storage unit; The camera unit captures an image of the driver at a current point in time, generating and storing a current image in the storage unit. The processing unit generates a local detection result using the local object detection model. Based on multiple target historical images taken within a predetermined time interval in these historical images, and the current image, it generates multiple local historical joint probabilities corresponding to the target historical images and a local current joint probability corresponding to the current image. Based on these local historical joint probabilities and the local current joint probability, the processing unit generates a local probability density function related to the probability distribution of the driving violation occurring in the current image and the target historical images. Based on the local probability density function and the baseline truth probability density function, the processing unit generates an error value. The electronic device determines whether the error value is less than a preset threshold. When the error value is determined to be less than the preset threshold, the electronic device generates a detection result indicating that the local detection result is credible.
7. The electronic device according to claim 6, characterized in that, It also includes a communication unit electrically connected to the processing unit. The communication unit is connected to a remote server via a communication network. The remote server stores a remote object detection model for determining whether an image contains multiple feature objects. The processing unit determines whether correction is needed. When correction is needed, the electronic device transmits the target historical images and the current image to the remote server via the communication network. The remote server uses the remote object detection model to generate a remote detection result and, based on the target historical images and the current image, generates a remote probability density function related to the probability distribution of the driving violation occurring in the current image and the target historical images. It also calculates multiple real-time parameters that conform to the remote probability density function and transmits them back via the communication network. The electronic device updates the baseline truth probability density function based on these real-time parameters. When correction is not needed, the processing unit generates the local historical joint probability and the current joint probability.
8. The electronic device according to claim 7, characterized in that, When the electronic device determines that the current time point has reached the preset period, the electronic device determines that correction is required.
9. The electronic device according to claim 6, characterized in that, The processing unit inputs the current image into the local object detection model, which outputs multiple probability values corresponding to the individual characteristics of the objects when they meet the driving violation criteria. The processing unit then multiplies these individual probability values to generate the local current joint probability corresponding to the current image.
10. The electronic device according to claim 7, characterized in that, The processing unit inputs the target historical image into the local object detection model. The local object detection model outputs multiple probability values for each of the feature objects in each frame of the target historical image when they meet the driving violation behavior. Then, the processing unit multiplies the individual probability values in each frame to generate multiple local historical joint probabilities for each frame of the target historical image.