Abstract: Characteristic data of a relevant component is extracted from a component image captured by camera, and the extracted characteristic data is compared with reference characteristic data of multiple types of component registered in a database to distinguish a type of component for which the two sets of data match most closely. As a result of this, a power spectrum image is created by applying an FFT for each scanning line to a lead presence candidate region corresponding to a candidate for a lead being present in a component image distinguished as an odd form component. Further, an HLAC characteristic is calculated from the power spectrum image in order to extract a periodic characteristic of the power spectrum image, and a connector component with leads is distinguished from an odd form component by learning a characteristic amount of the LAC characteristic using an SVM.

Abstract: An electronic component picked up and held by a mounting head of a component mounter is moved to an imaging position above a component imaging camera, the tip of the lead of the electronic component is imaged from below by the camera to acquire a lead image, and the lead image is inputted into an image recognition device configured from a neural network or the like and imaged processed so as to output the center position of the tip of the lead. Here, whenever an image processing error occurs, the lead image for which the image processing error occurred is displayed on a display device, and an operator specifies the center position of the tip of the lead in the lead image display on the display device and the specified center position is inputted into the image recognition device.

Abstract: An electronic component picked up and held by a mounting head of a component mounter is moved to an imaging position above a component imaging camera, the tip of the lead of the electronic component is imaged from below by the camera to acquire a lead image, and the lead image is inputted into an image recognition device configured from a neural network or the like and imaged processed so as to output the center position of the tip of the lead. Here, whenever an image processing error occurs, the lead image for which the image processing error occurred is displayed on a display device, and an operator specifies the center position of the tip of the lead in the lead image display on the display device and the specified center position is inputted into the image recognition device.

Abstract: When recognizing leads by processing an image captured by a camera of a component provided with a single row or multiple rows of leads of the same size and shape at a regular pitch, waveform analysis of a change pattern of brightness (pixel value) along multiple lines set in a vertical, a horizontal, or a diagonal direction of the image is performed so as to identify a region in which the brightness changes cyclically as a lead recognition target region in which there is a possibility of a lead existing. Then, image recognition of leads is performed in the lead recognition target region.

Abstract: The object detection device 1 detects an object being recognized (such as a pedestrian) in a frame image 42, and identifies an area 42a where a detected object which is detected in the frame image 42 is present. A frame image 43 is input after the frame image 42. The object detection device 1 detects the object being recognized in the frame image 43, and identifies an area 43a where a detected object which is detected in the frame image 43 is present. When a distance from center coordinates 42b of the area 42a to center coordinates 43b of the area 43a is smaller than a reference distance, the object detection device 1 determines that the detected object which is detected in the frame image 43 is identical to the detected object which is detected in the frame image 42.

Abstract: Characteristic data of a relevant component is extracted from a component image captured by camera, and the extracted characteristic data is compared with reference characteristic data of multiple types of component registered in a database to distinguish a type of component for which the two sets of data match most closely. As a result of this, a power spectrum image is created by applying an FFT for each scanning line to a lead presence candidate region corresponding to a candidate for a lead being present in a component image distinguished as an odd form component. Further, an HLAC characteristic is calculated from the power spectrum image in order to extract a periodic characteristic of the power spectrum image, and a connector component with leads is distinguished from an odd form component by learning a characteristic amount of the LAC characteristic using an SVM.

Abstract: The object detection device 1 detects an object being recognized (such as a pedestrian) in a frame image 42, and identifies an area 42a where a detected object which is detected in the frame image 42 is present. A frame image 43 is input after the frame image 42. The object detection device 1 detects the object being recognized in the frame image 43, and identifies an area 43a where a detected object which is detected in the frame image 43 is present. When a distance from center coordinates 42b of the area 42a to center coordinates 43b of the area 43a is smaller than a reference distance, the object detection device 1 determines that the detected object which is detected in the frame image 43 is identical to the detected object which is detected in the frame image 42.

Abstract: When recognizing leads by processing an image captured by a camera of a component provided with a single row or multiple rows of leads of the same size and shape at a regular pitch, waveform analysis of a change pattern of brightness (pixel value) along multiple lines set in a vertical, a horizontal, or a diagonal direction of the image is performed so as to identify a region in which the brightness changes cyclically as a lead recognition target region in which there is a possibility of a lead existing. Then, image recognition of leads is performed in the lead recognition target region.

Abstract: An object detection device detects an object being recognized (such as a pedestrian) in a frame image, and identifies an area where a detected object which is detected in the frame image is present. A frame image is input after the frame image. The object detection device detects the object being recognized in the frame image, and identifies an area where a detected object which is detected in the frame image is present. When a distance from center coordinates of the area to center coordinates of the area is smaller than a reference distance, the object detection device determines that the detected object which is detected in the frame image is identical to the detected object which is detected in the frame image.

Abstract: [Problem] To provide a low-cost, high-precision three-dimensional shape measuring device using vibration-resistant phase shift digital holography.

Abstract: A data processor includes an obtainment part that obtains an image of a person's face, a creation part that creates a face direction map in which face images of the person facing respective directions are arranged, based on the image of the person's face obtained by the obtainment part, and a determination part that determines movement of the person's face, based on the face direction map and a moving image of the person's face obtained by the obtainment part.

Abstract: An object detection device detects an object being recognized (such as a pedestrian) in a frame image, and identifies an area where a detected object which is detected in the frame image is present. A frame image is input after the frame image. The object detection device detects the object being recognized in the frame image, and identifies an area where a detected object which is detected in the frame image is present. When a distance from center coordinates of the area to center coordinates of the area smaller than a reference distance, the object detection device determines that the detected object which is detected in the frame image identical to the detected object which is detected in the frame image.

Abstract: [Problem] To provide a low-cost, high-precision three-dimensional shape measuring device using vibration-resistant phase shift digital holography.

Abstract: A data processor includes an obtainment part that obtains an image of a person's face, a creation part that creates a face direction map in which face images of the person facing respective directions are arranged, based on the image of the person's face obtained by the obtainment part, and a determination part that determines movement of the person's face, based on the face direction map and a moving image of the person's face obtained by the obtainment part.

Abstract: A hybrid biometric authentication device that includes a plurality of feature extraction portions that respectively extract, from biometric information, a plurality of types of feature information that have mutually distinct properties, a feature combining portion that generates hybrid feature information by combining the plurality of types of feature information, a feature information storage portion that stores the hybrid feature information, a plurality of similarity computation portions that separates each of new information and registered information into the plurality of types of feature information, and compute a plurality of degrees of similarity between separate sets of a same type of the feature information, and a determination portion that, based on the plurality of degrees of similarity, determines whether the biometric information on which the new information is based and the biometric information on which the registered information is based have been acquired from a same person.

Abstract: A hybrid biometric authentication device that includes a plurality of feature extraction portions that respectively extract, from biometric information, a plurality of types of feature information that have mutually distinct properties, a feature combining portion that generates hybrid feature information by combining the plurality of types of feature information, a feature information storage portion that stores the hybrid feature information, a plurality of similarity computation portions that separates each of new information and registered information into the plurality of types of feature information, and compute a plurality of degrees of similarity between separate sets of a same type of the feature information, and a determination portion that, based on the plurality of degrees of similarity, determines whether the biometric information on which the new information is based and the biometric information on which the registered information is based have been acquired from a same person.

Abstract: A defect inspection apparatus for inspecting an object to be inspected for a defect by processing an image taken from the object, includes: neural networks provided respectively for individual defect types to be classified; a learning unit which makes the neural networks learn based on the corresponding defect types to be classified; and a defect detection unit which classifies and detects defect types using the neural networks that have learned.

Abstract: An overlapped portion of a plurality of partial images of a fingerprint that are input from a fingerprint input device (7) is corrected to be reconfigured at a reconfiguration process section (106). The characteristics are extracted from the input partial images at the fingerprint characteristic information extracting section (110) and the extracted characteristic information is stored in an input characteristic information memory. A verification process section executes DP comparison (dynamic programming method) between the characteristic information stored in the input characteristic information memory and the characteristic information stored in the registered characteristic information memory. The distance value obtained by the DP comparison is compared to a threshold that is previously set. If the distance value is smaller than the threshold, it is judged the fingerprint is identified and if the distance value is larger than the threshold, it is judged that the fingerprint is not identified.

Abstract: The present invention provides technologies for narrowing down fingerprints images, which will be used for identity determination, without extracting auxiliary information such as a fingerprint pattern, thereby shortening the time required for fingerprint verification. The device according to the present invention includes a storage means 26 for storing characteristic information extracted from fingerprint images of registered persons, and a collection means 10 for collecting a fingerprint image. First, a extraction means 14 extracts characteristic information from the fingerprint image collected by the collection means 10. Next, a narrowing means 16 calculates a similarity between part of the characteristic information of the person to be identified, which was extracted by the extraction means 14, and corresponding part of the characteristic information of each of the registered persons, which is stored in the storage means 26.

Abstract: A defect inspection apparatus for inspecting an object to be inspected for a defect by processing an image taken from the object, includes: neural networks provided respectively for individual defect types to be classified; a learning unit which makes the neural networks learn based on the corresponding defect types to be classified; and a defect detection unit which classifies and detects defect types using the neural networks that have learned.