Sensor device and parameter setting method
A sensor device includes an array sensor in which a plurality of visible or non-visible light imaging elements is arranged in one or two dimensional manner, and an image processing unit that performs image processing using a parameter on which an instruction is given for the image signal obtained by imaging in the array sensor. In this case, a class is identified for an object detected from the image signal obtained by the imaging in the array sensor, the parameter to be used in the image processing is selected on the basis of the identified class, and a processing setting for the image processing unit is performed with the selected parameter.
1. A sensor device comprising:
an array sensor having a plurality of visible or non-visible imaging elements in one or two dimensional manner;
a first processor configured to perform image processing using a parameter on which an instruction is given, for an image signal obtained by imaging in the array sensor, and output the image signal as the image signal to be used for image recognition; and
a second processor configured to identify a class for an object detected from the image signal obtained by the imaging in the array sensor, select a parameter from a set of selectable parameters defining an image quality adjustment value to be used in the image processing of the image signal on a basis of the identified class, and perform a processing setting for the first processor, using the selected parameter,
wherein each selectable parameter is associated with a respective confidence rate, and the respective confidence rate associated with the selected parameter is a lowest respective confidence rate that is above a predefined confidence threshold for an object detection.
2. The sensor device according to claim 1 , further comprising:
an output interface configured to output the image signal processed by the first processor to an external device.
3. The sensor device according to claim 2 , wherein
the output interface also transmits information regarding class identification for the image signal to be output.
4. The sensor device according to claim 1 , wherein
the second processor performs
object region recognition processing of detecting a candidate object region of an object to be detected in one frame of the image signal,
class identification processing of identifying the class of the object in the object region detected in the object region recognition processing, and
parameter selection processing of performing parameter selection on a basis of an identification result of the class identification processing and performing the processing setting for the first processor.
5. The sensor device according to claim 1 , wherein
the second processor performs
object region recognition processing of detecting a candidate object region of an object to be detected in one frame of the image signal,
class identification processing of identifying the class of the object in the object region detected in the object region recognition processing,
processing of determining whether or not a target class is present among the classes identified in the class identification processing, and
parameter selection processing of performing parameter selection on a basis of a class recognition result for the target class, and performing the processing setting for the first processor in a case where the target class is present.
6. The sensor device according to claim 5 , wherein
a plurality of the target classes is settable, and
in a case where objects of a plurality of target classes are present in the one frame, the parameter selection processing is performed on a basis of selected one target class.
7. The sensor device according to claim 6 , wherein
a bounding box surrounding the object is calculated for the object region corresponding to the target class, and
the one target class is selected using an area of the bounding box in the case where a plurality of target classes is present.
8. The sensor device according to claim 1 , wherein
the second processor selects a parameter set including a plurality of parameters set for each class on a basis of the identified class.
9. The sensor device according to claim 8 , wherein
the parameter set is a set of a plurality of parameters obtained by deep learning using an image of the object corresponding to each class.
10. The sensor device according to claim 1 , wherein
the second processor identifies the class for a frame serving as a keyframe in the image signal obtained by the imaging in the array sensor, and selects the parameter to be used in the image processing on a basis of the identified class.
11. The sensor device according to claim 10 , wherein
the keyframe is a frame at predetermined time intervals.
12. The sensor device according to claim 10 , wherein
the keyframe is a frame at timing based on a command from an external device.
13. The sensor device according to claim 1 , wherein
the first processor performs, as the image processing for the image signal, at least one of color correction, gamma correction, color gradation processing, gain processing, contour enhancement processing, data compression processing, frame rate conversion, resolution conversion, aspect ratio conversion, contrast adjustment processing, sharpness adjustment processing, gray level adjustment processing, or sampling rate change processing.
14. The sensor device according to claim 1 , further comprising:
an output interface configured to output any or all of the image signal processed by the first processor, information of the identified class, the number of detected objects, and information of presence or absence of a target class in response to a request of an external device.
15. The sensor device according to claim 1 , wherein
the second processor is configured to set the confidence threshold for all or some of parameters to be used in the image processing of the first processor or imaging processing regarding the imaging by the array sensor, and enable the processing using the parameter set on a basis of the threshold to be performed.
16. The sensor device according to claim 15 , wherein
the second processor sets the confidence threshold according to the class of the object detected from the image signal.
17. The sensor device according to claim 15 , wherein
the second processor sets the confidence threshold on a basis of learning processing for the image signal.
18. The sensor device according to claim 15 , wherein
the second processor sets the confidence threshold so as to obtain a predetermined rate as a rate of certainty of object detection.
19. A parameter setting method in a sensor device including an array sensor having a plurality of visible or non-visible imaging elements in one or two dimensional manner, and a first processor configured to perform image processing using a parameter on which an instruction is given, for an image signal obtained by imaging in the array sensor, and output the image signal as the image signal for image recognition, the parameter setting method comprising:
identifying a class for an object detected from the image signal obtained by the imaging in the array sensor;
selecting a parameter defining an image quality adjustment value to be used in the image processing of the image signal on a basis of the identified class; and
performing a processing setting for the first processor, using the selected parameter,
wherein each selectable parameter is associated with a respective confidence rate, and the respective confidence rate associated with the selected parameter is a lowest respective confidence rate that is above a predefined confidence threshold.
20. The parameter setting method according to claim 19 , further comprising:
performing threshold setting processing of setting the confidence threshold for all or some of parameters to be used in the image processing of the first processor or imaging processing regarding the imaging by the array sensor; and
enabling the processing using the parameter changed on a basis of the confidence threshold to be performed.