Object detection system, autonomous vehicle using the same, and object detection method thereof

An autonomous vehicle comprising:

• an object detection system comprising:
• a first type of sensor for generating a first sensor data;
• a second type of sensor for generating a second sensor data;
• and
• a processor coupled to the first type of sensor and the second type of sensor and configured at least for:
• processing the first sensor data by using a first plurality of object detection algorithms to generate first preliminary detection results which correspond to the first type of sensor;
• processing the second sensor data by using a second plurality of object detection algorithms to generate second preliminary detection results which correspond to the second type of sensor;
• applying a parameter detection algorithm which comprises a plurality of environmental parameters for each of the first preliminary detection results and the second preliminary detection results to generate a plurality of confidence values with each confidence value corresponding to a different environmental parameter of the plurality of environmental parameters, wherein the plurality of confidence values are generated according to the first sensor data and the second sensor data;
• and
• determining a detected object based on characteristics of the first type of sensor, characteristics of the second type of sensor, relationships among the first preliminary detection results and the second preliminary detection results, and the plurality of confident values, comprising:
• determining the detected object from multiple bounding boxes according to the plurality of confidence values, wherein the plurality of confidence values are corresponded to the multiple bounding boxes respectively,
• wherein the processor is further configured for:
• fusing the first preliminary detection results and the second preliminary detection results to generate a fused preliminary detection results;
• paring two bounding boxes;
• and
• performing an overlap and clutter analysis from the fused preliminary detection results by determining whether two bounding boxes are overlapping above a predetermined overlapping threshold and are separated above a predetermined distance threshold,
• wherein the paired bounding boxes are determined as being either independent or dependent based on at least a spatial distance between the paired bounding boxes, depth information of the bounding boxes, and detected classes of the bounding boxes.

An object detection method used by an autonomous vehicle, the method comprising:

• generating, by using a first type of sensor, a first sensor data;
• generating, by using a second type of sensor, a second sensor data;
• processing the first sensor data by using a first plurality of object detection algorithms to generate first preliminary detection results which correspond to the first type of sensor;
• processing the second sensor data by using a second plurality of object detection algorithms to generate second preliminary detection results which correspond to the second type of sensor;
• applying a parameter detection algorithm which comprises a plurality of environmental parameters for each of the first preliminary detection results and the second preliminary detection results to generate a plurality of confidence values with each confidence value corresponding to a different environmental parameter of the plurality of environmental parameters, wherein the plurality of confidence values are generated according to the first sensor data and the second sensor data;
• and
• determining a detected object based on characteristics of the first type of sensor, characteristics of the second type of sensor, relationships among the first preliminary detection results and the second preliminary detection results, and the plurality of confident values, comprising:
• determining the detected object from multiple bounding boxes according to the plurality of confidence values, wherein the plurality of confidence values are corresponded to the multiple bounding boxes respectively,
• the object detection method further comprising:
• fusing the first preliminary detection results and the second detection results to generate a fused preliminary detection results;
• paring two bounding boxes;
• and
• performing an overlap and clutter analysis from the fused preliminary detection results by determining whether two bounding boxes are overlapping above a predetermined overlapping threshold and are separated above a predetermined distance threshold,
• wherein the paired bounding boxes are determined as being either independent or dependent based on at least a spatial distance between the paired bounding boxes, depth information of the bounding boxes, and detected classes of the bounding boxes.