An apparatus for identifying the state of a signal light includes a processor configured to input time series images into a first classifier to detect object regions each including a vehicle equipped with a signal light in the respective images, the first classifier having been trained to detect the vehicle; chronologically input characteristics obtained from pixel values of the object regions detected in the respective images into a second classifier to calculate confidence scores of possible candidate states of the signal light of the vehicle, the second classifier having a recursive structure or performing a convolution operation in a temporal direction; and identify the state of the signal light, based on the preceding state of the signal light, information indicating whether transitions between the candidate states of the signal light are allowed, and the confidence scores of the respective candidate states.

Systems, methods, tangible non-transitory computer-readable media, and devices for associating objects are provided. For example, the disclosed technology can receive sensor data associated with the detection of objects over time. An association dataset can be generated and can include information associated with object detections of the objects at a most recent time interval and object tracks of the objects at time intervals in the past. A subset of the association dataset including the object detections that satisfy some association subset criteria can be determined. Association scores for the object detections in the subset of the association dataset can be determined. Further, the object detections can be associated with the object tracks based on the association scores for each of the object detections in the subset of the association dataset that satisfy some association criteria.

An apparatus for identifying the state of an object includes a processor configured to input, every time obtaining an image from a camera, the image into a first classifier to detect, for each of one or more predetermined objects represented in the image, an object region including the object; determine a predicted object region in a subsequent image to be obtained from the camera for an object whose position in the subsequent image is predictable; and input characteristics into a second classifier to identify the state of an object involving time-varying changes in outward appearance. When the object has a predicted object region, the characteristics are obtained from pixel values of the predicted object region in the subsequent image. On the other hand, when the object does not have a predicted object region, the characteristics are obtained from pixel values of the object region detected from the subsequent image.

In a solid state imaging device as an embodiment, an analog-to-digital converter unit converts, in a first period, a first pixel signal into a digital signal, performs, in a determination period after the first period, the comparison of a second pixel signal with the reference signal set to a predetermined threshold, and converts, in a second period after the determination period, the second pixel signal at a gain in accordance with a result of the comparison performed in the determination period into a digital signal. Until the reference signal reaches the threshold from the first period, the reference signal generation unit changes the reference signal without changing a direction of change of the reference signal with respect to the lapse of time.

A navigation system may include a processor programmed to analyze a first image to identify a non-semantic road feature; identify a first image location, in the first image, of one point associated with the non-semantic road feature; analyze a second image to identify a representation of the non-semantic road feature in the second image; identify a second image location, in the second image, of the one point associated with the non-semantic road feature; determine, based on a difference between the first and second image locations and based on motion information for a vehicle between a capture of the first image and a capture of the second image, three-dimensional coordinates for the one point associated with the non-semantic road feature; and send the three-dimensional coordinates for the one point associated with the non-semantic road feature to a server for use in updating a road navigation model.

A system may include a processor configured to receive a first image captured during a drive of a first vehicle along a road segment and receive a second image captured during a drive of the second vehicle along the road segment; analyze the first and second images to identify representations of objects; analyze the first and second images to determine position indicators for each of the objects relative to the road segment; correlate the position indicators for each of the objects, wherein the correlating includes determining refined positions of each object based on the determined position indicators; and generate, based on the refined positions of objects belonging to a particular predetermined category of objects, a map including representations of the refined positions of one or more of the objects that belong to the particular predetermined category of objects.

In various examples, a deep neural network (DNN) is trained to accurately predict, in deployment, distances to objects and obstacles using image data alone. The DNN may be trained with ground truth data that is generated and encoded using sensor data from any number of depth predicting sensors, such as, without limitation, RADAR sensors, LIDAR sensors, and/or SONAR sensors. Camera adaptation algorithms may be used in various embodiments to adapt the DNN for use with image data generated by cameras with varying parameters—such as varying fields of view. In some examples, a post-processing safety bounds operation may be executed on the predictions of the DNN to ensure that the predictions fall within a safety-permissible range.

An attached object detection apparatus includes a setting part, a calculator, and a detector. A setting part sets a plurality of divided regions within a captured image captured by an image capturing apparatus. A calculator calculates, for each of the plurality of divided regions and based on luminance of pixels in the divided region, a representative luminance value and an amount of luminance dispersion. A detector detects, as an attached region that has an attached object, a divided region among the plurality of divided regions that satisfies the following conditions: i) a first difference is equal to or smaller than a first predetermined difference, ii) a second difference is equal to or smaller than a second predetermined difference, and iii) the representative luminance value of the current divided region is equal to or smaller than a first predetermined value.

A method and apparatus are provided for optimizing one or more object detection parameters used by an autonomous vehicle to detect objects in images. The autonomous vehicle may capture the images using one or more sensors. The autonomous vehicle may then determine object labels and their corresponding object label parameters for the detected objects. The captured images and the object label parameters may be communicated to an object identification server. The object identification server may request that one or more reviewers identify objects in the captured images. The object identification server may then compare the identification of objects by reviewers with the identification of objects by the autonomous vehicle. Depending on the results of the comparison, the object identification server may recommend or perform the optimization of one or more of the object detection parameters.

Computerized methods are performable by a driver assistance system while the host vehicle is moving. The driver assistance system includes a camera connectible to a processor. First and second image frames are captured from the field of view of the camera. Corresponding image points of the road are tracked from the first image frame to the second image frame. Image motion between the corresponding image points of the road is processed to detect a hazard in the road. The corresponding image points are determined to be of a moving shadow cast on the road to avoid a false positive detection of a hazard in the road or the corresponding image points are determined not to be of a moving shadow cast on the road to verify detection of a hazard in the road.