Object recognition device, object recognition system, and object recognition method

Abstract

It accurately recognizes the internal structure of an object. [Solution] The object recognition device includes: a work information acquisition unit that acquires a perspective image capturing the internal structure of a target object and three-dimensional information showing the external shape of the target object; a standard bone model acquisition unit that acquires a standard bone model including the three-dimensional information of the object; a pseudo-perspective image generation unit that generates a pseudo-perspective image, which is pseudo-two-dimensional information simulating the internal structure of the target object; a comparison unit that calculates three-dimensional coordinates including depth based on the three-dimensional information acquired by the work information acquisition unit and compares them with pseudo-two-dimensional information simulating the internal structure of the target object; and an output unit that outputs the parameters when the difference between the internal structure of the target object and the pseudo-internal structure simulating the internal structure of the target object falls below a predetermined threshold.

Claims

1. A workpiece information acquisition unit acquires a transparent image capturing the internal structure of the target object and three-dimensional information showing the external shape of the target object. A standard bone model acquisition unit acquires a standard bone model that includes three-dimensional information of an object, which is pre-generated information from a plurality of objects having similar attributes to the aforementioned target object. A pseudo-perspective image generation unit generates a pseudo-perspective image, which is pseudo-two-dimensional information that simulates the internal structure of the target object, by changing parameters based on the standard bone model. A comparison unit calculates three-dimensional coordinates including depth based on the three-dimensional information acquired by the work information acquisition unit for the internal structure of the target object shown in the perspective image acquired by the work information acquisition unit, and compares the internal structure of the target object with pseudo-two-dimensional information that simulates the internal structure of the target object identified from the pseudo-perspective image generated by the pseudo-perspective image generation unit. If, as a result of the comparison by the comparison unit, the difference between the internal structure of the target object and the pseudo-internal structure that simulates the internal structure of the target object falls below a predetermined threshold, the output unit outputs the parameters used to generate the pseudo-perspective image. An object recognition device equipped with the following features.

2. The parameters are adjusted until the difference between the internal structure of the target object and the simulated internal structure that mimics the target object falls below a predetermined threshold, and the generation of the simulated perspective image and the comparison between the internal structure of the target object and the simulated internal structure that mimics the target object are repeated. The object recognition device according to claim 1.

3. The parameters modified by the pseudo-transparency image generation unit include parameters for moving, rotating, and deforming the standard bone model. The object recognition device according to claim 1.

4. The aforementioned standard bone model is voxel information in normal grid units in three-dimensional space. The object recognition device according to claim 1.

5. The aforementioned fluoroscopic image is an X-ray image obtained by irradiating the target object with X-rays and visualizing the transmitted X-rays. The object recognition device according to claim 1.

6. The aforementioned object is meat, The aforementioned internal structure includes two or more bones, each containing at least one joint. The aforementioned parameters include information for moving the joints. The object recognition device according to claim 1.

7. A workpiece information acquisition unit acquires a transparent image capturing the internal structure of the target object and three-dimensional information showing the external shape of the target object. A standard bone model acquisition unit acquires a standard bone model that includes three-dimensional information of an object, which is pre-generated information from a plurality of objects having similar attributes to the aforementioned target object. A pseudo-perspective image generation unit generates a pseudo-perspective image, which is pseudo-two-dimensional information that simulates the internal structure of the target object, by changing parameters based on the standard bone model. A comparison unit calculates three-dimensional coordinates including depth based on the three-dimensional information acquired by the work information acquisition unit for the internal structure of the target object shown in the perspective image acquired by the work information acquisition unit, and compares the internal structure of the target object with pseudo-two-dimensional information that simulates the internal structure of the target object identified from the pseudo-perspective image generated by the pseudo-perspective image generation unit. If, as a result of the comparison by the comparison unit, the difference between the internal structure of the target object and the pseudo-internal structure that simulates the internal structure of the target object falls below a predetermined threshold, the output unit outputs the parameters used to generate the pseudo-perspective image. The parameters are adjusted until the difference between the internal structure of the target object and the simulated internal structure that mimics the target object falls below a predetermined threshold, and the generation of the simulated perspective image and the comparison between the internal structure of the target object and the simulated internal structure that mimics the target object are repeated. Equipped with, The aforementioned object is meat, The aforementioned internal structure includes two or more bones, each containing at least one joint. The aforementioned parameters include information for moving the joints. The pseudo-transparency image generation unit generates the pseudo-transparency image of the internal structure formed by the connection of multiple bones and the pseudo-transparency image of each individual bone separately and independently. First, the comparison unit compares the pseudo-transparency image of the internal structure formed by the connection of multiple bones with the transparency image of the internal structure formed by the connection of multiple bones. If, as a result of the comparison by the comparison unit, the difference between the internal structure of the target object and the pseudo-internal structure that simulates the target object is not below a predetermined threshold, second, the comparison unit compares the pseudo-transparency image of the internal structure of each bone with the transparency image of the internal structure of each bone. Object recognition device.

8. The aforementioned internal structure has a structure in which the first bone, the second bone, and the third bone are connected in series. When the comparison unit compares the pseudo-fluoroscopic image of the internal structure of each bone with the fluoroscopic image of the internal structure of each bone, it first compares the second bone, and then compares the first bone and the third bone located at both ends of the second bone. The object recognition device according to claim 7.

9. An object recognition device according to any one of claims 1 to 8, An imaging device for capturing the aforementioned fluoroscopic image, A three-dimensional information acquisition device that acquires the aforementioned three-dimensional information, An object recognition system equipped with the following features.

10. A processing unit located downstream of the object recognition device on the transport path of the target object, which processes the target object based on the information output by the object recognition device. The object recognition system according to claim 9, comprising:

11. A workpiece information acquisition step involves acquiring a transparent image capturing the internal structure of the target object and three-dimensional information showing the external shape of the target object. A standard bone model acquisition step involves acquiring a standard bone model that includes three-dimensional information of an object, which is pre-generated information from a plurality of objects having similar attributes to the aforementioned target object. A pseudo-transparency image generation step generates a pseudo-transparency image, which is pseudo-two-dimensional information that simulates the internal structure of the target object, by changing parameters based on the standard bone model, A comparison step which involves calculating three-dimensional coordinates including depth based on the three-dimensional information obtained in the work information acquisition step for the internal structure of the target object shown in the perspective image obtained in the work information acquisition step, and comparing the internal structure of the target object with pseudo-two-dimensional information that simulates the internal structure of the target object identified from the pseudo-perspective image generated in the pseudo-perspective image generation step, If, as a result of the comparison in the comparison step, the difference between the internal structure of the target object and the pseudo-internal structure that simulates the internal structure of the target object falls below a predetermined threshold, an output step is performed to output the parameters used to generate the pseudo-perspective image. A method for recognizing harmful objects.

Images