OBSTACLE DETECTION DEVICE, OBSTACLE DETECTION SYSTEM INCLUDING THE SAME, AND OBSTACLE DETECTION METHOD

• Main Authors: KOKIDO TOMOYOSHI, TAKEHARA NARIAKI, FUJIYOSHI HIROKI
• Format: Patent
• Language: English; Japanese
• Published: 01.11.2021
• Subjects: CALCULATING; COMPUTING; COUNTING; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS; SIGNALLING; TRAFFIC CONTROL SYSTEMS

To provide an obstacle detection device and an obstacle detection system improved in detection accuracy with respect to an obstacle without additionally providing a sensor other than a camera.SOLUTION: An obstacle detection device 100 includes an optical flow calculation section 3, a road surface detection section 4, and an object detection section 6. Erroneous detection by optical flow processing can be eliminated by road surface processing and object detection processing based on learning. Approach feature points present in a road surface area are eliminated from approach feature points extracted by the optical flow calculation section 3, so as to eliminate approach feature points related to a crack 17 of a road and a white line 18 or the like. Besides, whether an obstacle or not is determined by object detection processing based on learning concerning approach feature points present in an area other than a road surface area. Thus, approach feature points related to regularly installed poles and pylons or the like are not determined as an obstacle.SELECTED DRAWING: Figure 1 【課題】カメラ以外の他のセンサを追加して設けることなく、障害物に対する検知精度を向上させた障害物検知装置及び障害物検知システムを得る。【解決手段】障害物検知装置１００は、オプティカルフロー演算部３、路面検知部４、及び物体検知部６を備えているので、オプティカルフロー処理による誤検知を路面検知処理と学習に基づく物体検知処理によって除去することができる。オプティカルフロー演算部３で抽出された接近特徴点から路面領域に存在する接近特徴点を除去することにより、道路のひび割れ１７、白線１８等に係る接近特徴点は除去される。また、路面領域以外の領域に存在する接近特徴点については、学習に基づく物体検知処理により障害物か否かが判別される。これにより、規則正しく設置されたポール及びパイロン等に係る接近特徴点は障害物でないと判別される。【選択図】図１